1 //===-- ObjectFileMachO.cpp -------------------------------------*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 // C Includes 11 // C++ Includes 12 // Other libraries and framework includes 13 #include "llvm/ADT/StringRef.h" 14 15 // Project includes 16 #include "Plugins/Process/Utility/RegisterContextDarwin_arm.h" 17 #include "Plugins/Process/Utility/RegisterContextDarwin_arm64.h" 18 #include "Plugins/Process/Utility/RegisterContextDarwin_i386.h" 19 #include "Plugins/Process/Utility/RegisterContextDarwin_x86_64.h" 20 #include "lldb/Core/Debugger.h" 21 #include "lldb/Core/FileSpecList.h" 22 #include "lldb/Core/Module.h" 23 #include "lldb/Core/ModuleSpec.h" 24 #include "lldb/Core/PluginManager.h" 25 #include "lldb/Core/RangeMap.h" 26 #include "lldb/Core/RegisterValue.h" 27 #include "lldb/Core/Section.h" 28 #include "lldb/Core/StreamFile.h" 29 #include "lldb/Host/Host.h" 30 #include "lldb/Symbol/DWARFCallFrameInfo.h" 31 #include "lldb/Symbol/ObjectFile.h" 32 #include "lldb/Target/DynamicLoader.h" 33 #include "lldb/Target/MemoryRegionInfo.h" 34 #include "lldb/Target/Platform.h" 35 #include "lldb/Target/Process.h" 36 #include "lldb/Target/SectionLoadList.h" 37 #include "lldb/Target/Target.h" 38 #include "lldb/Target/Thread.h" 39 #include "lldb/Target/ThreadList.h" 40 #include "lldb/Utility/ArchSpec.h" 41 #include "lldb/Utility/DataBuffer.h" 42 #include "lldb/Utility/FileSpec.h" 43 #include "lldb/Utility/Log.h" 44 #include "lldb/Utility/Status.h" 45 #include "lldb/Utility/StreamString.h" 46 #include "lldb/Utility/Timer.h" 47 #include "lldb/Utility/UUID.h" 48 49 #include "lldb/Utility/SafeMachO.h" 50 51 #include "llvm/Support/MemoryBuffer.h" 52 53 #include "ObjectFileMachO.h" 54 55 #if defined(__APPLE__) && \ 56 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 57 // GetLLDBSharedCacheUUID() needs to call dlsym() 58 #include <dlfcn.h> 59 #endif 60 61 #ifndef __APPLE__ 62 #include "Utility/UuidCompatibility.h" 63 #else 64 #include <uuid/uuid.h> 65 #endif 66 67 #define THUMB_ADDRESS_BIT_MASK 0xfffffffffffffffeull 68 using namespace lldb; 69 using namespace lldb_private; 70 using namespace llvm::MachO; 71 72 // Some structure definitions needed for parsing the dyld shared cache files 73 // found on iOS devices. 74 75 struct lldb_copy_dyld_cache_header_v1 { 76 char magic[16]; // e.g. "dyld_v0 i386", "dyld_v1 armv7", etc. 77 uint32_t mappingOffset; // file offset to first dyld_cache_mapping_info 78 uint32_t mappingCount; // number of dyld_cache_mapping_info entries 79 uint32_t imagesOffset; 80 uint32_t imagesCount; 81 uint64_t dyldBaseAddress; 82 uint64_t codeSignatureOffset; 83 uint64_t codeSignatureSize; 84 uint64_t slideInfoOffset; 85 uint64_t slideInfoSize; 86 uint64_t localSymbolsOffset; 87 uint64_t localSymbolsSize; 88 uint8_t uuid[16]; // v1 and above, also recorded in dyld_all_image_infos v13 89 // and later 90 }; 91 92 struct lldb_copy_dyld_cache_mapping_info { 93 uint64_t address; 94 uint64_t size; 95 uint64_t fileOffset; 96 uint32_t maxProt; 97 uint32_t initProt; 98 }; 99 100 struct lldb_copy_dyld_cache_local_symbols_info { 101 uint32_t nlistOffset; 102 uint32_t nlistCount; 103 uint32_t stringsOffset; 104 uint32_t stringsSize; 105 uint32_t entriesOffset; 106 uint32_t entriesCount; 107 }; 108 struct lldb_copy_dyld_cache_local_symbols_entry { 109 uint32_t dylibOffset; 110 uint32_t nlistStartIndex; 111 uint32_t nlistCount; 112 }; 113 114 class RegisterContextDarwin_x86_64_Mach : public RegisterContextDarwin_x86_64 { 115 public: 116 RegisterContextDarwin_x86_64_Mach(lldb_private::Thread &thread, 117 const DataExtractor &data) 118 : RegisterContextDarwin_x86_64(thread, 0) { 119 SetRegisterDataFrom_LC_THREAD(data); 120 } 121 122 void InvalidateAllRegisters() override { 123 // Do nothing... registers are always valid... 124 } 125 126 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 127 lldb::offset_t offset = 0; 128 SetError(GPRRegSet, Read, -1); 129 SetError(FPURegSet, Read, -1); 130 SetError(EXCRegSet, Read, -1); 131 bool done = false; 132 133 while (!done) { 134 int flavor = data.GetU32(&offset); 135 if (flavor == 0) 136 done = true; 137 else { 138 uint32_t i; 139 uint32_t count = data.GetU32(&offset); 140 switch (flavor) { 141 case GPRRegSet: 142 for (i = 0; i < count; ++i) 143 (&gpr.rax)[i] = data.GetU64(&offset); 144 SetError(GPRRegSet, Read, 0); 145 done = true; 146 147 break; 148 case FPURegSet: 149 // TODO: fill in FPU regs.... 150 // SetError (FPURegSet, Read, -1); 151 done = true; 152 153 break; 154 case EXCRegSet: 155 exc.trapno = data.GetU32(&offset); 156 exc.err = data.GetU32(&offset); 157 exc.faultvaddr = data.GetU64(&offset); 158 SetError(EXCRegSet, Read, 0); 159 done = true; 160 break; 161 case 7: 162 case 8: 163 case 9: 164 // fancy flavors that encapsulate of the above flavors... 165 break; 166 167 default: 168 done = true; 169 break; 170 } 171 } 172 } 173 } 174 175 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 176 const char *alt_name, size_t reg_byte_size, 177 Stream &data) { 178 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 179 if (reg_info == NULL) 180 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 181 if (reg_info) { 182 lldb_private::RegisterValue reg_value; 183 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 184 if (reg_info->byte_size >= reg_byte_size) 185 data.Write(reg_value.GetBytes(), reg_byte_size); 186 else { 187 data.Write(reg_value.GetBytes(), reg_info->byte_size); 188 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 189 ++i) 190 data.PutChar(0); 191 } 192 return reg_byte_size; 193 } 194 } 195 // Just write zeros if all else fails 196 for (size_t i = 0; i < reg_byte_size; ++i) 197 data.PutChar(0); 198 return reg_byte_size; 199 } 200 201 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 202 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 203 if (reg_ctx_sp) { 204 RegisterContext *reg_ctx = reg_ctx_sp.get(); 205 206 data.PutHex32(GPRRegSet); // Flavor 207 data.PutHex32(GPRWordCount); 208 WriteRegister(reg_ctx, "rax", NULL, 8, data); 209 WriteRegister(reg_ctx, "rbx", NULL, 8, data); 210 WriteRegister(reg_ctx, "rcx", NULL, 8, data); 211 WriteRegister(reg_ctx, "rdx", NULL, 8, data); 212 WriteRegister(reg_ctx, "rdi", NULL, 8, data); 213 WriteRegister(reg_ctx, "rsi", NULL, 8, data); 214 WriteRegister(reg_ctx, "rbp", NULL, 8, data); 215 WriteRegister(reg_ctx, "rsp", NULL, 8, data); 216 WriteRegister(reg_ctx, "r8", NULL, 8, data); 217 WriteRegister(reg_ctx, "r9", NULL, 8, data); 218 WriteRegister(reg_ctx, "r10", NULL, 8, data); 219 WriteRegister(reg_ctx, "r11", NULL, 8, data); 220 WriteRegister(reg_ctx, "r12", NULL, 8, data); 221 WriteRegister(reg_ctx, "r13", NULL, 8, data); 222 WriteRegister(reg_ctx, "r14", NULL, 8, data); 223 WriteRegister(reg_ctx, "r15", NULL, 8, data); 224 WriteRegister(reg_ctx, "rip", NULL, 8, data); 225 WriteRegister(reg_ctx, "rflags", NULL, 8, data); 226 WriteRegister(reg_ctx, "cs", NULL, 8, data); 227 WriteRegister(reg_ctx, "fs", NULL, 8, data); 228 WriteRegister(reg_ctx, "gs", NULL, 8, data); 229 230 // // Write out the FPU registers 231 // const size_t fpu_byte_size = sizeof(FPU); 232 // size_t bytes_written = 0; 233 // data.PutHex32 (FPURegSet); 234 // data.PutHex32 (fpu_byte_size/sizeof(uint64_t)); 235 // bytes_written += data.PutHex32(0); // uint32_t pad[0] 236 // bytes_written += data.PutHex32(0); // uint32_t pad[1] 237 // bytes_written += WriteRegister (reg_ctx, "fcw", "fctrl", 2, 238 // data); // uint16_t fcw; // "fctrl" 239 // bytes_written += WriteRegister (reg_ctx, "fsw" , "fstat", 2, 240 // data); // uint16_t fsw; // "fstat" 241 // bytes_written += WriteRegister (reg_ctx, "ftw" , "ftag", 1, 242 // data); // uint8_t ftw; // "ftag" 243 // bytes_written += data.PutHex8 (0); // uint8_t pad1; 244 // bytes_written += WriteRegister (reg_ctx, "fop" , NULL, 2, 245 // data); // uint16_t fop; // "fop" 246 // bytes_written += WriteRegister (reg_ctx, "fioff", "ip", 4, 247 // data); // uint32_t ip; // "fioff" 248 // bytes_written += WriteRegister (reg_ctx, "fiseg", NULL, 2, 249 // data); // uint16_t cs; // "fiseg" 250 // bytes_written += data.PutHex16 (0); // uint16_t pad2; 251 // bytes_written += WriteRegister (reg_ctx, "dp", "fooff" , 4, 252 // data); // uint32_t dp; // "fooff" 253 // bytes_written += WriteRegister (reg_ctx, "foseg", NULL, 2, 254 // data); // uint16_t ds; // "foseg" 255 // bytes_written += data.PutHex16 (0); // uint16_t pad3; 256 // bytes_written += WriteRegister (reg_ctx, "mxcsr", NULL, 4, 257 // data); // uint32_t mxcsr; 258 // bytes_written += WriteRegister (reg_ctx, "mxcsrmask", NULL, 259 // 4, data);// uint32_t mxcsrmask; 260 // bytes_written += WriteRegister (reg_ctx, "stmm0", NULL, 261 // sizeof(MMSReg), data); 262 // bytes_written += WriteRegister (reg_ctx, "stmm1", NULL, 263 // sizeof(MMSReg), data); 264 // bytes_written += WriteRegister (reg_ctx, "stmm2", NULL, 265 // sizeof(MMSReg), data); 266 // bytes_written += WriteRegister (reg_ctx, "stmm3", NULL, 267 // sizeof(MMSReg), data); 268 // bytes_written += WriteRegister (reg_ctx, "stmm4", NULL, 269 // sizeof(MMSReg), data); 270 // bytes_written += WriteRegister (reg_ctx, "stmm5", NULL, 271 // sizeof(MMSReg), data); 272 // bytes_written += WriteRegister (reg_ctx, "stmm6", NULL, 273 // sizeof(MMSReg), data); 274 // bytes_written += WriteRegister (reg_ctx, "stmm7", NULL, 275 // sizeof(MMSReg), data); 276 // bytes_written += WriteRegister (reg_ctx, "xmm0" , NULL, 277 // sizeof(XMMReg), data); 278 // bytes_written += WriteRegister (reg_ctx, "xmm1" , NULL, 279 // sizeof(XMMReg), data); 280 // bytes_written += WriteRegister (reg_ctx, "xmm2" , NULL, 281 // sizeof(XMMReg), data); 282 // bytes_written += WriteRegister (reg_ctx, "xmm3" , NULL, 283 // sizeof(XMMReg), data); 284 // bytes_written += WriteRegister (reg_ctx, "xmm4" , NULL, 285 // sizeof(XMMReg), data); 286 // bytes_written += WriteRegister (reg_ctx, "xmm5" , NULL, 287 // sizeof(XMMReg), data); 288 // bytes_written += WriteRegister (reg_ctx, "xmm6" , NULL, 289 // sizeof(XMMReg), data); 290 // bytes_written += WriteRegister (reg_ctx, "xmm7" , NULL, 291 // sizeof(XMMReg), data); 292 // bytes_written += WriteRegister (reg_ctx, "xmm8" , NULL, 293 // sizeof(XMMReg), data); 294 // bytes_written += WriteRegister (reg_ctx, "xmm9" , NULL, 295 // sizeof(XMMReg), data); 296 // bytes_written += WriteRegister (reg_ctx, "xmm10", NULL, 297 // sizeof(XMMReg), data); 298 // bytes_written += WriteRegister (reg_ctx, "xmm11", NULL, 299 // sizeof(XMMReg), data); 300 // bytes_written += WriteRegister (reg_ctx, "xmm12", NULL, 301 // sizeof(XMMReg), data); 302 // bytes_written += WriteRegister (reg_ctx, "xmm13", NULL, 303 // sizeof(XMMReg), data); 304 // bytes_written += WriteRegister (reg_ctx, "xmm14", NULL, 305 // sizeof(XMMReg), data); 306 // bytes_written += WriteRegister (reg_ctx, "xmm15", NULL, 307 // sizeof(XMMReg), data); 308 // 309 // // Fill rest with zeros 310 // for (size_t i=0, n = fpu_byte_size - bytes_written; i<n; ++ 311 // i) 312 // data.PutChar(0); 313 314 // Write out the EXC registers 315 data.PutHex32(EXCRegSet); 316 data.PutHex32(EXCWordCount); 317 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 318 WriteRegister(reg_ctx, "err", NULL, 4, data); 319 WriteRegister(reg_ctx, "faultvaddr", NULL, 8, data); 320 return true; 321 } 322 return false; 323 } 324 325 protected: 326 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 327 328 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 329 330 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 331 332 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 333 return 0; 334 } 335 336 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 337 return 0; 338 } 339 340 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 341 return 0; 342 } 343 }; 344 345 class RegisterContextDarwin_i386_Mach : public RegisterContextDarwin_i386 { 346 public: 347 RegisterContextDarwin_i386_Mach(lldb_private::Thread &thread, 348 const DataExtractor &data) 349 : RegisterContextDarwin_i386(thread, 0) { 350 SetRegisterDataFrom_LC_THREAD(data); 351 } 352 353 void InvalidateAllRegisters() override { 354 // Do nothing... registers are always valid... 355 } 356 357 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 358 lldb::offset_t offset = 0; 359 SetError(GPRRegSet, Read, -1); 360 SetError(FPURegSet, Read, -1); 361 SetError(EXCRegSet, Read, -1); 362 bool done = false; 363 364 while (!done) { 365 int flavor = data.GetU32(&offset); 366 if (flavor == 0) 367 done = true; 368 else { 369 uint32_t i; 370 uint32_t count = data.GetU32(&offset); 371 switch (flavor) { 372 case GPRRegSet: 373 for (i = 0; i < count; ++i) 374 (&gpr.eax)[i] = data.GetU32(&offset); 375 SetError(GPRRegSet, Read, 0); 376 done = true; 377 378 break; 379 case FPURegSet: 380 // TODO: fill in FPU regs.... 381 // SetError (FPURegSet, Read, -1); 382 done = true; 383 384 break; 385 case EXCRegSet: 386 exc.trapno = data.GetU32(&offset); 387 exc.err = data.GetU32(&offset); 388 exc.faultvaddr = data.GetU32(&offset); 389 SetError(EXCRegSet, Read, 0); 390 done = true; 391 break; 392 case 7: 393 case 8: 394 case 9: 395 // fancy flavors that encapsulate of the above flavors... 396 break; 397 398 default: 399 done = true; 400 break; 401 } 402 } 403 } 404 } 405 406 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 407 const char *alt_name, size_t reg_byte_size, 408 Stream &data) { 409 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 410 if (reg_info == NULL) 411 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 412 if (reg_info) { 413 lldb_private::RegisterValue reg_value; 414 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 415 if (reg_info->byte_size >= reg_byte_size) 416 data.Write(reg_value.GetBytes(), reg_byte_size); 417 else { 418 data.Write(reg_value.GetBytes(), reg_info->byte_size); 419 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 420 ++i) 421 data.PutChar(0); 422 } 423 return reg_byte_size; 424 } 425 } 426 // Just write zeros if all else fails 427 for (size_t i = 0; i < reg_byte_size; ++i) 428 data.PutChar(0); 429 return reg_byte_size; 430 } 431 432 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 433 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 434 if (reg_ctx_sp) { 435 RegisterContext *reg_ctx = reg_ctx_sp.get(); 436 437 data.PutHex32(GPRRegSet); // Flavor 438 data.PutHex32(GPRWordCount); 439 WriteRegister(reg_ctx, "eax", NULL, 4, data); 440 WriteRegister(reg_ctx, "ebx", NULL, 4, data); 441 WriteRegister(reg_ctx, "ecx", NULL, 4, data); 442 WriteRegister(reg_ctx, "edx", NULL, 4, data); 443 WriteRegister(reg_ctx, "edi", NULL, 4, data); 444 WriteRegister(reg_ctx, "esi", NULL, 4, data); 445 WriteRegister(reg_ctx, "ebp", NULL, 4, data); 446 WriteRegister(reg_ctx, "esp", NULL, 4, data); 447 WriteRegister(reg_ctx, "ss", NULL, 4, data); 448 WriteRegister(reg_ctx, "eflags", NULL, 4, data); 449 WriteRegister(reg_ctx, "eip", NULL, 4, data); 450 WriteRegister(reg_ctx, "cs", NULL, 4, data); 451 WriteRegister(reg_ctx, "ds", NULL, 4, data); 452 WriteRegister(reg_ctx, "es", NULL, 4, data); 453 WriteRegister(reg_ctx, "fs", NULL, 4, data); 454 WriteRegister(reg_ctx, "gs", NULL, 4, data); 455 456 // Write out the EXC registers 457 data.PutHex32(EXCRegSet); 458 data.PutHex32(EXCWordCount); 459 WriteRegister(reg_ctx, "trapno", NULL, 4, data); 460 WriteRegister(reg_ctx, "err", NULL, 4, data); 461 WriteRegister(reg_ctx, "faultvaddr", NULL, 4, data); 462 return true; 463 } 464 return false; 465 } 466 467 protected: 468 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return 0; } 469 470 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return 0; } 471 472 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return 0; } 473 474 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 475 return 0; 476 } 477 478 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 479 return 0; 480 } 481 482 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 483 return 0; 484 } 485 }; 486 487 class RegisterContextDarwin_arm_Mach : public RegisterContextDarwin_arm { 488 public: 489 RegisterContextDarwin_arm_Mach(lldb_private::Thread &thread, 490 const DataExtractor &data) 491 : RegisterContextDarwin_arm(thread, 0) { 492 SetRegisterDataFrom_LC_THREAD(data); 493 } 494 495 void InvalidateAllRegisters() override { 496 // Do nothing... registers are always valid... 497 } 498 499 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 500 lldb::offset_t offset = 0; 501 SetError(GPRRegSet, Read, -1); 502 SetError(FPURegSet, Read, -1); 503 SetError(EXCRegSet, Read, -1); 504 bool done = false; 505 506 while (!done) { 507 int flavor = data.GetU32(&offset); 508 uint32_t count = data.GetU32(&offset); 509 lldb::offset_t next_thread_state = offset + (count * 4); 510 switch (flavor) { 511 case GPRAltRegSet: 512 case GPRRegSet: 513 for (uint32_t i = 0; i < count; ++i) { 514 gpr.r[i] = data.GetU32(&offset); 515 } 516 517 // Note that gpr.cpsr is also copied by the above loop; this loop 518 // technically extends one element past the end of the gpr.r[] array. 519 520 SetError(GPRRegSet, Read, 0); 521 offset = next_thread_state; 522 break; 523 524 case FPURegSet: { 525 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.floats.s[0]; 526 const int fpu_reg_buf_size = sizeof(fpu.floats); 527 if (data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 528 fpu_reg_buf) == fpu_reg_buf_size) { 529 offset += fpu_reg_buf_size; 530 fpu.fpscr = data.GetU32(&offset); 531 SetError(FPURegSet, Read, 0); 532 } else { 533 done = true; 534 } 535 } 536 offset = next_thread_state; 537 break; 538 539 case EXCRegSet: 540 if (count == 3) { 541 exc.exception = data.GetU32(&offset); 542 exc.fsr = data.GetU32(&offset); 543 exc.far = data.GetU32(&offset); 544 SetError(EXCRegSet, Read, 0); 545 } 546 done = true; 547 offset = next_thread_state; 548 break; 549 550 // Unknown register set flavor, stop trying to parse. 551 default: 552 done = true; 553 } 554 } 555 } 556 557 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 558 const char *alt_name, size_t reg_byte_size, 559 Stream &data) { 560 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 561 if (reg_info == NULL) 562 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 563 if (reg_info) { 564 lldb_private::RegisterValue reg_value; 565 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 566 if (reg_info->byte_size >= reg_byte_size) 567 data.Write(reg_value.GetBytes(), reg_byte_size); 568 else { 569 data.Write(reg_value.GetBytes(), reg_info->byte_size); 570 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 571 ++i) 572 data.PutChar(0); 573 } 574 return reg_byte_size; 575 } 576 } 577 // Just write zeros if all else fails 578 for (size_t i = 0; i < reg_byte_size; ++i) 579 data.PutChar(0); 580 return reg_byte_size; 581 } 582 583 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 584 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 585 if (reg_ctx_sp) { 586 RegisterContext *reg_ctx = reg_ctx_sp.get(); 587 588 data.PutHex32(GPRRegSet); // Flavor 589 data.PutHex32(GPRWordCount); 590 WriteRegister(reg_ctx, "r0", NULL, 4, data); 591 WriteRegister(reg_ctx, "r1", NULL, 4, data); 592 WriteRegister(reg_ctx, "r2", NULL, 4, data); 593 WriteRegister(reg_ctx, "r3", NULL, 4, data); 594 WriteRegister(reg_ctx, "r4", NULL, 4, data); 595 WriteRegister(reg_ctx, "r5", NULL, 4, data); 596 WriteRegister(reg_ctx, "r6", NULL, 4, data); 597 WriteRegister(reg_ctx, "r7", NULL, 4, data); 598 WriteRegister(reg_ctx, "r8", NULL, 4, data); 599 WriteRegister(reg_ctx, "r9", NULL, 4, data); 600 WriteRegister(reg_ctx, "r10", NULL, 4, data); 601 WriteRegister(reg_ctx, "r11", NULL, 4, data); 602 WriteRegister(reg_ctx, "r12", NULL, 4, data); 603 WriteRegister(reg_ctx, "sp", NULL, 4, data); 604 WriteRegister(reg_ctx, "lr", NULL, 4, data); 605 WriteRegister(reg_ctx, "pc", NULL, 4, data); 606 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 607 608 // Write out the EXC registers 609 // data.PutHex32 (EXCRegSet); 610 // data.PutHex32 (EXCWordCount); 611 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 612 // WriteRegister (reg_ctx, "fsr", NULL, 4, data); 613 // WriteRegister (reg_ctx, "far", NULL, 4, data); 614 return true; 615 } 616 return false; 617 } 618 619 protected: 620 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 621 622 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 623 624 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 625 626 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 627 628 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 629 return 0; 630 } 631 632 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 633 return 0; 634 } 635 636 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 637 return 0; 638 } 639 640 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 641 return -1; 642 } 643 }; 644 645 class RegisterContextDarwin_arm64_Mach : public RegisterContextDarwin_arm64 { 646 public: 647 RegisterContextDarwin_arm64_Mach(lldb_private::Thread &thread, 648 const DataExtractor &data) 649 : RegisterContextDarwin_arm64(thread, 0) { 650 SetRegisterDataFrom_LC_THREAD(data); 651 } 652 653 void InvalidateAllRegisters() override { 654 // Do nothing... registers are always valid... 655 } 656 657 void SetRegisterDataFrom_LC_THREAD(const DataExtractor &data) { 658 lldb::offset_t offset = 0; 659 SetError(GPRRegSet, Read, -1); 660 SetError(FPURegSet, Read, -1); 661 SetError(EXCRegSet, Read, -1); 662 bool done = false; 663 while (!done) { 664 int flavor = data.GetU32(&offset); 665 uint32_t count = data.GetU32(&offset); 666 lldb::offset_t next_thread_state = offset + (count * 4); 667 switch (flavor) { 668 case GPRRegSet: 669 // x0-x29 + fp + lr + sp + pc (== 33 64-bit registers) plus cpsr (1 670 // 32-bit register) 671 if (count >= (33 * 2) + 1) { 672 for (uint32_t i = 0; i < 29; ++i) 673 gpr.x[i] = data.GetU64(&offset); 674 gpr.fp = data.GetU64(&offset); 675 gpr.lr = data.GetU64(&offset); 676 gpr.sp = data.GetU64(&offset); 677 gpr.pc = data.GetU64(&offset); 678 gpr.cpsr = data.GetU32(&offset); 679 SetError(GPRRegSet, Read, 0); 680 } 681 offset = next_thread_state; 682 break; 683 case FPURegSet: { 684 uint8_t *fpu_reg_buf = (uint8_t *)&fpu.v[0]; 685 const int fpu_reg_buf_size = sizeof(fpu); 686 if (fpu_reg_buf_size == count * sizeof(uint32_t) && 687 data.ExtractBytes(offset, fpu_reg_buf_size, eByteOrderLittle, 688 fpu_reg_buf) == fpu_reg_buf_size) { 689 SetError(FPURegSet, Read, 0); 690 } else { 691 done = true; 692 } 693 } 694 offset = next_thread_state; 695 break; 696 case EXCRegSet: 697 if (count == 4) { 698 exc.far = data.GetU64(&offset); 699 exc.esr = data.GetU32(&offset); 700 exc.exception = data.GetU32(&offset); 701 SetError(EXCRegSet, Read, 0); 702 } 703 offset = next_thread_state; 704 break; 705 default: 706 done = true; 707 break; 708 } 709 } 710 } 711 712 static size_t WriteRegister(RegisterContext *reg_ctx, const char *name, 713 const char *alt_name, size_t reg_byte_size, 714 Stream &data) { 715 const RegisterInfo *reg_info = reg_ctx->GetRegisterInfoByName(name); 716 if (reg_info == NULL) 717 reg_info = reg_ctx->GetRegisterInfoByName(alt_name); 718 if (reg_info) { 719 lldb_private::RegisterValue reg_value; 720 if (reg_ctx->ReadRegister(reg_info, reg_value)) { 721 if (reg_info->byte_size >= reg_byte_size) 722 data.Write(reg_value.GetBytes(), reg_byte_size); 723 else { 724 data.Write(reg_value.GetBytes(), reg_info->byte_size); 725 for (size_t i = 0, n = reg_byte_size - reg_info->byte_size; i < n; 726 ++i) 727 data.PutChar(0); 728 } 729 return reg_byte_size; 730 } 731 } 732 // Just write zeros if all else fails 733 for (size_t i = 0; i < reg_byte_size; ++i) 734 data.PutChar(0); 735 return reg_byte_size; 736 } 737 738 static bool Create_LC_THREAD(Thread *thread, Stream &data) { 739 RegisterContextSP reg_ctx_sp(thread->GetRegisterContext()); 740 if (reg_ctx_sp) { 741 RegisterContext *reg_ctx = reg_ctx_sp.get(); 742 743 data.PutHex32(GPRRegSet); // Flavor 744 data.PutHex32(GPRWordCount); 745 WriteRegister(reg_ctx, "x0", NULL, 8, data); 746 WriteRegister(reg_ctx, "x1", NULL, 8, data); 747 WriteRegister(reg_ctx, "x2", NULL, 8, data); 748 WriteRegister(reg_ctx, "x3", NULL, 8, data); 749 WriteRegister(reg_ctx, "x4", NULL, 8, data); 750 WriteRegister(reg_ctx, "x5", NULL, 8, data); 751 WriteRegister(reg_ctx, "x6", NULL, 8, data); 752 WriteRegister(reg_ctx, "x7", NULL, 8, data); 753 WriteRegister(reg_ctx, "x8", NULL, 8, data); 754 WriteRegister(reg_ctx, "x9", NULL, 8, data); 755 WriteRegister(reg_ctx, "x10", NULL, 8, data); 756 WriteRegister(reg_ctx, "x11", NULL, 8, data); 757 WriteRegister(reg_ctx, "x12", NULL, 8, data); 758 WriteRegister(reg_ctx, "x13", NULL, 8, data); 759 WriteRegister(reg_ctx, "x14", NULL, 8, data); 760 WriteRegister(reg_ctx, "x15", NULL, 8, data); 761 WriteRegister(reg_ctx, "x16", NULL, 8, data); 762 WriteRegister(reg_ctx, "x17", NULL, 8, data); 763 WriteRegister(reg_ctx, "x18", NULL, 8, data); 764 WriteRegister(reg_ctx, "x19", NULL, 8, data); 765 WriteRegister(reg_ctx, "x20", NULL, 8, data); 766 WriteRegister(reg_ctx, "x21", NULL, 8, data); 767 WriteRegister(reg_ctx, "x22", NULL, 8, data); 768 WriteRegister(reg_ctx, "x23", NULL, 8, data); 769 WriteRegister(reg_ctx, "x24", NULL, 8, data); 770 WriteRegister(reg_ctx, "x25", NULL, 8, data); 771 WriteRegister(reg_ctx, "x26", NULL, 8, data); 772 WriteRegister(reg_ctx, "x27", NULL, 8, data); 773 WriteRegister(reg_ctx, "x28", NULL, 8, data); 774 WriteRegister(reg_ctx, "fp", NULL, 8, data); 775 WriteRegister(reg_ctx, "lr", NULL, 8, data); 776 WriteRegister(reg_ctx, "sp", NULL, 8, data); 777 WriteRegister(reg_ctx, "pc", NULL, 8, data); 778 WriteRegister(reg_ctx, "cpsr", NULL, 4, data); 779 780 // Write out the EXC registers 781 // data.PutHex32 (EXCRegSet); 782 // data.PutHex32 (EXCWordCount); 783 // WriteRegister (reg_ctx, "far", NULL, 8, data); 784 // WriteRegister (reg_ctx, "esr", NULL, 4, data); 785 // WriteRegister (reg_ctx, "exception", NULL, 4, data); 786 return true; 787 } 788 return false; 789 } 790 791 protected: 792 int DoReadGPR(lldb::tid_t tid, int flavor, GPR &gpr) override { return -1; } 793 794 int DoReadFPU(lldb::tid_t tid, int flavor, FPU &fpu) override { return -1; } 795 796 int DoReadEXC(lldb::tid_t tid, int flavor, EXC &exc) override { return -1; } 797 798 int DoReadDBG(lldb::tid_t tid, int flavor, DBG &dbg) override { return -1; } 799 800 int DoWriteGPR(lldb::tid_t tid, int flavor, const GPR &gpr) override { 801 return 0; 802 } 803 804 int DoWriteFPU(lldb::tid_t tid, int flavor, const FPU &fpu) override { 805 return 0; 806 } 807 808 int DoWriteEXC(lldb::tid_t tid, int flavor, const EXC &exc) override { 809 return 0; 810 } 811 812 int DoWriteDBG(lldb::tid_t tid, int flavor, const DBG &dbg) override { 813 return -1; 814 } 815 }; 816 817 static uint32_t MachHeaderSizeFromMagic(uint32_t magic) { 818 switch (magic) { 819 case MH_MAGIC: 820 case MH_CIGAM: 821 return sizeof(struct mach_header); 822 823 case MH_MAGIC_64: 824 case MH_CIGAM_64: 825 return sizeof(struct mach_header_64); 826 break; 827 828 default: 829 break; 830 } 831 return 0; 832 } 833 834 #define MACHO_NLIST_ARM_SYMBOL_IS_THUMB 0x0008 835 836 void ObjectFileMachO::Initialize() { 837 PluginManager::RegisterPlugin( 838 GetPluginNameStatic(), GetPluginDescriptionStatic(), CreateInstance, 839 CreateMemoryInstance, GetModuleSpecifications, SaveCore); 840 } 841 842 void ObjectFileMachO::Terminate() { 843 PluginManager::UnregisterPlugin(CreateInstance); 844 } 845 846 lldb_private::ConstString ObjectFileMachO::GetPluginNameStatic() { 847 static ConstString g_name("mach-o"); 848 return g_name; 849 } 850 851 const char *ObjectFileMachO::GetPluginDescriptionStatic() { 852 return "Mach-o object file reader (32 and 64 bit)"; 853 } 854 855 ObjectFile *ObjectFileMachO::CreateInstance(const lldb::ModuleSP &module_sp, 856 DataBufferSP &data_sp, 857 lldb::offset_t data_offset, 858 const FileSpec *file, 859 lldb::offset_t file_offset, 860 lldb::offset_t length) { 861 if (!data_sp) { 862 data_sp = MapFileData(*file, length, file_offset); 863 if (!data_sp) 864 return nullptr; 865 data_offset = 0; 866 } 867 868 if (!ObjectFileMachO::MagicBytesMatch(data_sp, data_offset, length)) 869 return nullptr; 870 871 // Update the data to contain the entire file if it doesn't already 872 if (data_sp->GetByteSize() < length) { 873 data_sp = MapFileData(*file, length, file_offset); 874 if (!data_sp) 875 return nullptr; 876 data_offset = 0; 877 } 878 auto objfile_ap = llvm::make_unique<ObjectFileMachO>( 879 module_sp, data_sp, data_offset, file, file_offset, length); 880 if (!objfile_ap || !objfile_ap->ParseHeader()) 881 return nullptr; 882 883 return objfile_ap.release(); 884 } 885 886 ObjectFile *ObjectFileMachO::CreateMemoryInstance( 887 const lldb::ModuleSP &module_sp, DataBufferSP &data_sp, 888 const ProcessSP &process_sp, lldb::addr_t header_addr) { 889 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 890 std::unique_ptr<ObjectFile> objfile_ap( 891 new ObjectFileMachO(module_sp, data_sp, process_sp, header_addr)); 892 if (objfile_ap.get() && objfile_ap->ParseHeader()) 893 return objfile_ap.release(); 894 } 895 return NULL; 896 } 897 898 size_t ObjectFileMachO::GetModuleSpecifications( 899 const lldb_private::FileSpec &file, lldb::DataBufferSP &data_sp, 900 lldb::offset_t data_offset, lldb::offset_t file_offset, 901 lldb::offset_t length, lldb_private::ModuleSpecList &specs) { 902 const size_t initial_count = specs.GetSize(); 903 904 if (ObjectFileMachO::MagicBytesMatch(data_sp, 0, data_sp->GetByteSize())) { 905 DataExtractor data; 906 data.SetData(data_sp); 907 llvm::MachO::mach_header header; 908 if (ParseHeader(data, &data_offset, header)) { 909 size_t header_and_load_cmds = 910 header.sizeofcmds + MachHeaderSizeFromMagic(header.magic); 911 if (header_and_load_cmds >= data_sp->GetByteSize()) { 912 data_sp = MapFileData(file, header_and_load_cmds, file_offset); 913 data.SetData(data_sp); 914 data_offset = MachHeaderSizeFromMagic(header.magic); 915 } 916 if (data_sp) { 917 ModuleSpec spec; 918 spec.GetFileSpec() = file; 919 spec.SetObjectOffset(file_offset); 920 spec.SetObjectSize(length); 921 922 if (GetArchitecture(header, data, data_offset, 923 spec.GetArchitecture())) { 924 if (spec.GetArchitecture().IsValid()) { 925 GetUUID(header, data, data_offset, spec.GetUUID()); 926 specs.Append(spec); 927 } 928 } 929 } 930 } 931 } 932 return specs.GetSize() - initial_count; 933 } 934 935 const ConstString &ObjectFileMachO::GetSegmentNameTEXT() { 936 static ConstString g_segment_name_TEXT("__TEXT"); 937 return g_segment_name_TEXT; 938 } 939 940 const ConstString &ObjectFileMachO::GetSegmentNameDATA() { 941 static ConstString g_segment_name_DATA("__DATA"); 942 return g_segment_name_DATA; 943 } 944 945 const ConstString &ObjectFileMachO::GetSegmentNameDATA_DIRTY() { 946 static ConstString g_segment_name("__DATA_DIRTY"); 947 return g_segment_name; 948 } 949 950 const ConstString &ObjectFileMachO::GetSegmentNameDATA_CONST() { 951 static ConstString g_segment_name("__DATA_CONST"); 952 return g_segment_name; 953 } 954 955 const ConstString &ObjectFileMachO::GetSegmentNameOBJC() { 956 static ConstString g_segment_name_OBJC("__OBJC"); 957 return g_segment_name_OBJC; 958 } 959 960 const ConstString &ObjectFileMachO::GetSegmentNameLINKEDIT() { 961 static ConstString g_section_name_LINKEDIT("__LINKEDIT"); 962 return g_section_name_LINKEDIT; 963 } 964 965 const ConstString &ObjectFileMachO::GetSectionNameEHFrame() { 966 static ConstString g_section_name_eh_frame("__eh_frame"); 967 return g_section_name_eh_frame; 968 } 969 970 bool ObjectFileMachO::MagicBytesMatch(DataBufferSP &data_sp, 971 lldb::addr_t data_offset, 972 lldb::addr_t data_length) { 973 DataExtractor data; 974 data.SetData(data_sp, data_offset, data_length); 975 lldb::offset_t offset = 0; 976 uint32_t magic = data.GetU32(&offset); 977 return MachHeaderSizeFromMagic(magic) != 0; 978 } 979 980 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 981 DataBufferSP &data_sp, 982 lldb::offset_t data_offset, 983 const FileSpec *file, 984 lldb::offset_t file_offset, 985 lldb::offset_t length) 986 : ObjectFile(module_sp, file, file_offset, length, data_sp, data_offset), 987 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 988 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 989 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 990 ::memset(&m_header, 0, sizeof(m_header)); 991 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 992 } 993 994 ObjectFileMachO::ObjectFileMachO(const lldb::ModuleSP &module_sp, 995 lldb::DataBufferSP &header_data_sp, 996 const lldb::ProcessSP &process_sp, 997 lldb::addr_t header_addr) 998 : ObjectFile(module_sp, process_sp, header_addr, header_data_sp), 999 m_mach_segments(), m_mach_sections(), m_entry_point_address(), 1000 m_thread_context_offsets(), m_thread_context_offsets_valid(false), 1001 m_reexported_dylibs(), m_allow_assembly_emulation_unwind_plans(true) { 1002 ::memset(&m_header, 0, sizeof(m_header)); 1003 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1004 } 1005 1006 bool ObjectFileMachO::ParseHeader(DataExtractor &data, 1007 lldb::offset_t *data_offset_ptr, 1008 llvm::MachO::mach_header &header) { 1009 data.SetByteOrder(endian::InlHostByteOrder()); 1010 // Leave magic in the original byte order 1011 header.magic = data.GetU32(data_offset_ptr); 1012 bool can_parse = false; 1013 bool is_64_bit = false; 1014 switch (header.magic) { 1015 case MH_MAGIC: 1016 data.SetByteOrder(endian::InlHostByteOrder()); 1017 data.SetAddressByteSize(4); 1018 can_parse = true; 1019 break; 1020 1021 case MH_MAGIC_64: 1022 data.SetByteOrder(endian::InlHostByteOrder()); 1023 data.SetAddressByteSize(8); 1024 can_parse = true; 1025 is_64_bit = true; 1026 break; 1027 1028 case MH_CIGAM: 1029 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1030 ? eByteOrderLittle 1031 : eByteOrderBig); 1032 data.SetAddressByteSize(4); 1033 can_parse = true; 1034 break; 1035 1036 case MH_CIGAM_64: 1037 data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1038 ? eByteOrderLittle 1039 : eByteOrderBig); 1040 data.SetAddressByteSize(8); 1041 is_64_bit = true; 1042 can_parse = true; 1043 break; 1044 1045 default: 1046 break; 1047 } 1048 1049 if (can_parse) { 1050 data.GetU32(data_offset_ptr, &header.cputype, 6); 1051 if (is_64_bit) 1052 *data_offset_ptr += 4; 1053 return true; 1054 } else { 1055 memset(&header, 0, sizeof(header)); 1056 } 1057 return false; 1058 } 1059 1060 bool ObjectFileMachO::ParseHeader() { 1061 ModuleSP module_sp(GetModule()); 1062 if (module_sp) { 1063 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1064 bool can_parse = false; 1065 lldb::offset_t offset = 0; 1066 m_data.SetByteOrder(endian::InlHostByteOrder()); 1067 // Leave magic in the original byte order 1068 m_header.magic = m_data.GetU32(&offset); 1069 switch (m_header.magic) { 1070 case MH_MAGIC: 1071 m_data.SetByteOrder(endian::InlHostByteOrder()); 1072 m_data.SetAddressByteSize(4); 1073 can_parse = true; 1074 break; 1075 1076 case MH_MAGIC_64: 1077 m_data.SetByteOrder(endian::InlHostByteOrder()); 1078 m_data.SetAddressByteSize(8); 1079 can_parse = true; 1080 break; 1081 1082 case MH_CIGAM: 1083 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1084 ? eByteOrderLittle 1085 : eByteOrderBig); 1086 m_data.SetAddressByteSize(4); 1087 can_parse = true; 1088 break; 1089 1090 case MH_CIGAM_64: 1091 m_data.SetByteOrder(endian::InlHostByteOrder() == eByteOrderBig 1092 ? eByteOrderLittle 1093 : eByteOrderBig); 1094 m_data.SetAddressByteSize(8); 1095 can_parse = true; 1096 break; 1097 1098 default: 1099 break; 1100 } 1101 1102 if (can_parse) { 1103 m_data.GetU32(&offset, &m_header.cputype, 6); 1104 1105 ArchSpec mach_arch; 1106 1107 if (GetArchitecture(mach_arch)) { 1108 // Check if the module has a required architecture 1109 const ArchSpec &module_arch = module_sp->GetArchitecture(); 1110 if (module_arch.IsValid() && !module_arch.IsCompatibleMatch(mach_arch)) 1111 return false; 1112 1113 if (SetModulesArchitecture(mach_arch)) { 1114 const size_t header_and_lc_size = 1115 m_header.sizeofcmds + MachHeaderSizeFromMagic(m_header.magic); 1116 if (m_data.GetByteSize() < header_and_lc_size) { 1117 DataBufferSP data_sp; 1118 ProcessSP process_sp(m_process_wp.lock()); 1119 if (process_sp) { 1120 data_sp = 1121 ReadMemory(process_sp, m_memory_addr, header_and_lc_size); 1122 } else { 1123 // Read in all only the load command data from the file on disk 1124 data_sp = MapFileData(m_file, header_and_lc_size, m_file_offset); 1125 if (data_sp->GetByteSize() != header_and_lc_size) 1126 return false; 1127 } 1128 if (data_sp) 1129 m_data.SetData(data_sp); 1130 } 1131 } 1132 return true; 1133 } 1134 } else { 1135 memset(&m_header, 0, sizeof(struct mach_header)); 1136 } 1137 } 1138 return false; 1139 } 1140 1141 ByteOrder ObjectFileMachO::GetByteOrder() const { 1142 return m_data.GetByteOrder(); 1143 } 1144 1145 bool ObjectFileMachO::IsExecutable() const { 1146 return m_header.filetype == MH_EXECUTE; 1147 } 1148 1149 uint32_t ObjectFileMachO::GetAddressByteSize() const { 1150 return m_data.GetAddressByteSize(); 1151 } 1152 1153 AddressClass ObjectFileMachO::GetAddressClass(lldb::addr_t file_addr) { 1154 Symtab *symtab = GetSymtab(); 1155 if (symtab) { 1156 Symbol *symbol = symtab->FindSymbolContainingFileAddress(file_addr); 1157 if (symbol) { 1158 if (symbol->ValueIsAddress()) { 1159 SectionSP section_sp(symbol->GetAddressRef().GetSection()); 1160 if (section_sp) { 1161 const lldb::SectionType section_type = section_sp->GetType(); 1162 switch (section_type) { 1163 case eSectionTypeInvalid: 1164 return eAddressClassUnknown; 1165 1166 case eSectionTypeCode: 1167 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1168 // For ARM we have a bit in the n_desc field of the symbol that 1169 // tells us ARM/Thumb which is bit 0x0008. 1170 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1171 return eAddressClassCodeAlternateISA; 1172 } 1173 return eAddressClassCode; 1174 1175 case eSectionTypeContainer: 1176 return eAddressClassUnknown; 1177 1178 case eSectionTypeData: 1179 case eSectionTypeDataCString: 1180 case eSectionTypeDataCStringPointers: 1181 case eSectionTypeDataSymbolAddress: 1182 case eSectionTypeData4: 1183 case eSectionTypeData8: 1184 case eSectionTypeData16: 1185 case eSectionTypeDataPointers: 1186 case eSectionTypeZeroFill: 1187 case eSectionTypeDataObjCMessageRefs: 1188 case eSectionTypeDataObjCCFStrings: 1189 case eSectionTypeGoSymtab: 1190 return eAddressClassData; 1191 1192 case eSectionTypeDebug: 1193 case eSectionTypeDWARFDebugAbbrev: 1194 case eSectionTypeDWARFDebugAddr: 1195 case eSectionTypeDWARFDebugAranges: 1196 case eSectionTypeDWARFDebugCuIndex: 1197 case eSectionTypeDWARFDebugFrame: 1198 case eSectionTypeDWARFDebugInfo: 1199 case eSectionTypeDWARFDebugLine: 1200 case eSectionTypeDWARFDebugLoc: 1201 case eSectionTypeDWARFDebugMacInfo: 1202 case eSectionTypeDWARFDebugMacro: 1203 case eSectionTypeDWARFDebugPubNames: 1204 case eSectionTypeDWARFDebugPubTypes: 1205 case eSectionTypeDWARFDebugRanges: 1206 case eSectionTypeDWARFDebugStr: 1207 case eSectionTypeDWARFDebugStrOffsets: 1208 case eSectionTypeDWARFDebugTypes: 1209 case eSectionTypeDWARFAppleNames: 1210 case eSectionTypeDWARFAppleTypes: 1211 case eSectionTypeDWARFAppleNamespaces: 1212 case eSectionTypeDWARFAppleObjC: 1213 case eSectionTypeDWARFGNUDebugAltLink: 1214 return eAddressClassDebug; 1215 1216 case eSectionTypeEHFrame: 1217 case eSectionTypeARMexidx: 1218 case eSectionTypeARMextab: 1219 case eSectionTypeCompactUnwind: 1220 return eAddressClassRuntime; 1221 1222 case eSectionTypeAbsoluteAddress: 1223 case eSectionTypeELFSymbolTable: 1224 case eSectionTypeELFDynamicSymbols: 1225 case eSectionTypeELFRelocationEntries: 1226 case eSectionTypeELFDynamicLinkInfo: 1227 case eSectionTypeOther: 1228 return eAddressClassUnknown; 1229 } 1230 } 1231 } 1232 1233 const SymbolType symbol_type = symbol->GetType(); 1234 switch (symbol_type) { 1235 case eSymbolTypeAny: 1236 return eAddressClassUnknown; 1237 case eSymbolTypeAbsolute: 1238 return eAddressClassUnknown; 1239 1240 case eSymbolTypeCode: 1241 case eSymbolTypeTrampoline: 1242 case eSymbolTypeResolver: 1243 if (m_header.cputype == llvm::MachO::CPU_TYPE_ARM) { 1244 // For ARM we have a bit in the n_desc field of the symbol that tells 1245 // us ARM/Thumb which is bit 0x0008. 1246 if (symbol->GetFlags() & MACHO_NLIST_ARM_SYMBOL_IS_THUMB) 1247 return eAddressClassCodeAlternateISA; 1248 } 1249 return eAddressClassCode; 1250 1251 case eSymbolTypeData: 1252 return eAddressClassData; 1253 case eSymbolTypeRuntime: 1254 return eAddressClassRuntime; 1255 case eSymbolTypeException: 1256 return eAddressClassRuntime; 1257 case eSymbolTypeSourceFile: 1258 return eAddressClassDebug; 1259 case eSymbolTypeHeaderFile: 1260 return eAddressClassDebug; 1261 case eSymbolTypeObjectFile: 1262 return eAddressClassDebug; 1263 case eSymbolTypeCommonBlock: 1264 return eAddressClassDebug; 1265 case eSymbolTypeBlock: 1266 return eAddressClassDebug; 1267 case eSymbolTypeLocal: 1268 return eAddressClassData; 1269 case eSymbolTypeParam: 1270 return eAddressClassData; 1271 case eSymbolTypeVariable: 1272 return eAddressClassData; 1273 case eSymbolTypeVariableType: 1274 return eAddressClassDebug; 1275 case eSymbolTypeLineEntry: 1276 return eAddressClassDebug; 1277 case eSymbolTypeLineHeader: 1278 return eAddressClassDebug; 1279 case eSymbolTypeScopeBegin: 1280 return eAddressClassDebug; 1281 case eSymbolTypeScopeEnd: 1282 return eAddressClassDebug; 1283 case eSymbolTypeAdditional: 1284 return eAddressClassUnknown; 1285 case eSymbolTypeCompiler: 1286 return eAddressClassDebug; 1287 case eSymbolTypeInstrumentation: 1288 return eAddressClassDebug; 1289 case eSymbolTypeUndefined: 1290 return eAddressClassUnknown; 1291 case eSymbolTypeObjCClass: 1292 return eAddressClassRuntime; 1293 case eSymbolTypeObjCMetaClass: 1294 return eAddressClassRuntime; 1295 case eSymbolTypeObjCIVar: 1296 return eAddressClassRuntime; 1297 case eSymbolTypeReExported: 1298 return eAddressClassRuntime; 1299 } 1300 } 1301 } 1302 return eAddressClassUnknown; 1303 } 1304 1305 Symtab *ObjectFileMachO::GetSymtab() { 1306 ModuleSP module_sp(GetModule()); 1307 if (module_sp) { 1308 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 1309 if (m_symtab_ap.get() == NULL) { 1310 m_symtab_ap.reset(new Symtab(this)); 1311 std::lock_guard<std::recursive_mutex> symtab_guard( 1312 m_symtab_ap->GetMutex()); 1313 ParseSymtab(); 1314 m_symtab_ap->Finalize(); 1315 } 1316 } 1317 return m_symtab_ap.get(); 1318 } 1319 1320 bool ObjectFileMachO::IsStripped() { 1321 if (m_dysymtab.cmd == 0) { 1322 ModuleSP module_sp(GetModule()); 1323 if (module_sp) { 1324 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1325 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1326 const lldb::offset_t load_cmd_offset = offset; 1327 1328 load_command lc; 1329 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 1330 break; 1331 if (lc.cmd == LC_DYSYMTAB) { 1332 m_dysymtab.cmd = lc.cmd; 1333 m_dysymtab.cmdsize = lc.cmdsize; 1334 if (m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1335 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2) == 1336 NULL) { 1337 // Clear m_dysymtab if we were unable to read all items from the 1338 // load command 1339 ::memset(&m_dysymtab, 0, sizeof(m_dysymtab)); 1340 } 1341 } 1342 offset = load_cmd_offset + lc.cmdsize; 1343 } 1344 } 1345 } 1346 if (m_dysymtab.cmd) 1347 return m_dysymtab.nlocalsym <= 1; 1348 return false; 1349 } 1350 1351 ObjectFileMachO::EncryptedFileRanges ObjectFileMachO::GetEncryptedFileRanges() { 1352 EncryptedFileRanges result; 1353 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1354 1355 encryption_info_command encryption_cmd; 1356 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1357 const lldb::offset_t load_cmd_offset = offset; 1358 if (m_data.GetU32(&offset, &encryption_cmd, 2) == NULL) 1359 break; 1360 1361 // LC_ENCRYPTION_INFO and LC_ENCRYPTION_INFO_64 have the same sizes for the 1362 // 3 fields we care about, so treat them the same. 1363 if (encryption_cmd.cmd == LC_ENCRYPTION_INFO || 1364 encryption_cmd.cmd == LC_ENCRYPTION_INFO_64) { 1365 if (m_data.GetU32(&offset, &encryption_cmd.cryptoff, 3)) { 1366 if (encryption_cmd.cryptid != 0) { 1367 EncryptedFileRanges::Entry entry; 1368 entry.SetRangeBase(encryption_cmd.cryptoff); 1369 entry.SetByteSize(encryption_cmd.cryptsize); 1370 result.Append(entry); 1371 } 1372 } 1373 } 1374 offset = load_cmd_offset + encryption_cmd.cmdsize; 1375 } 1376 1377 return result; 1378 } 1379 1380 void ObjectFileMachO::SanitizeSegmentCommand(segment_command_64 &seg_cmd, 1381 uint32_t cmd_idx) { 1382 if (m_length == 0 || seg_cmd.filesize == 0) 1383 return; 1384 1385 if (seg_cmd.fileoff > m_length) { 1386 // We have a load command that says it extends past the end of the file. 1387 // This is likely a corrupt file. We don't have any way to return an error 1388 // condition here (this method was likely invoked from something like 1389 // ObjectFile::GetSectionList()), so we just null out the section contents, 1390 // and dump a message to stdout. The most common case here is core file 1391 // debugging with a truncated file. 1392 const char *lc_segment_name = 1393 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1394 GetModule()->ReportWarning( 1395 "load command %u %s has a fileoff (0x%" PRIx64 1396 ") that extends beyond the end of the file (0x%" PRIx64 1397 "), ignoring this section", 1398 cmd_idx, lc_segment_name, seg_cmd.fileoff, m_length); 1399 1400 seg_cmd.fileoff = 0; 1401 seg_cmd.filesize = 0; 1402 } 1403 1404 if (seg_cmd.fileoff + seg_cmd.filesize > m_length) { 1405 // We have a load command that says it extends past the end of the file. 1406 // This is likely a corrupt file. We don't have any way to return an error 1407 // condition here (this method was likely invoked from something like 1408 // ObjectFile::GetSectionList()), so we just null out the section contents, 1409 // and dump a message to stdout. The most common case here is core file 1410 // debugging with a truncated file. 1411 const char *lc_segment_name = 1412 seg_cmd.cmd == LC_SEGMENT_64 ? "LC_SEGMENT_64" : "LC_SEGMENT"; 1413 GetModule()->ReportWarning( 1414 "load command %u %s has a fileoff + filesize (0x%" PRIx64 1415 ") that extends beyond the end of the file (0x%" PRIx64 1416 "), the segment will be truncated to match", 1417 cmd_idx, lc_segment_name, seg_cmd.fileoff + seg_cmd.filesize, m_length); 1418 1419 // Truncate the length 1420 seg_cmd.filesize = m_length - seg_cmd.fileoff; 1421 } 1422 } 1423 1424 static uint32_t GetSegmentPermissions(const segment_command_64 &seg_cmd) { 1425 uint32_t result = 0; 1426 if (seg_cmd.initprot & VM_PROT_READ) 1427 result |= ePermissionsReadable; 1428 if (seg_cmd.initprot & VM_PROT_WRITE) 1429 result |= ePermissionsWritable; 1430 if (seg_cmd.initprot & VM_PROT_EXECUTE) 1431 result |= ePermissionsExecutable; 1432 return result; 1433 } 1434 1435 static lldb::SectionType GetSectionType(uint32_t flags, 1436 ConstString section_name) { 1437 1438 if (flags & (S_ATTR_PURE_INSTRUCTIONS | S_ATTR_SOME_INSTRUCTIONS)) 1439 return eSectionTypeCode; 1440 1441 uint32_t mach_sect_type = flags & SECTION_TYPE; 1442 static ConstString g_sect_name_objc_data("__objc_data"); 1443 static ConstString g_sect_name_objc_msgrefs("__objc_msgrefs"); 1444 static ConstString g_sect_name_objc_selrefs("__objc_selrefs"); 1445 static ConstString g_sect_name_objc_classrefs("__objc_classrefs"); 1446 static ConstString g_sect_name_objc_superrefs("__objc_superrefs"); 1447 static ConstString g_sect_name_objc_const("__objc_const"); 1448 static ConstString g_sect_name_objc_classlist("__objc_classlist"); 1449 static ConstString g_sect_name_cfstring("__cfstring"); 1450 1451 static ConstString g_sect_name_dwarf_debug_abbrev("__debug_abbrev"); 1452 static ConstString g_sect_name_dwarf_debug_aranges("__debug_aranges"); 1453 static ConstString g_sect_name_dwarf_debug_frame("__debug_frame"); 1454 static ConstString g_sect_name_dwarf_debug_info("__debug_info"); 1455 static ConstString g_sect_name_dwarf_debug_line("__debug_line"); 1456 static ConstString g_sect_name_dwarf_debug_loc("__debug_loc"); 1457 static ConstString g_sect_name_dwarf_debug_macinfo("__debug_macinfo"); 1458 static ConstString g_sect_name_dwarf_debug_pubnames("__debug_pubnames"); 1459 static ConstString g_sect_name_dwarf_debug_pubtypes("__debug_pubtypes"); 1460 static ConstString g_sect_name_dwarf_debug_ranges("__debug_ranges"); 1461 static ConstString g_sect_name_dwarf_debug_str("__debug_str"); 1462 static ConstString g_sect_name_dwarf_debug_types("__debug_types"); 1463 static ConstString g_sect_name_dwarf_apple_names("__apple_names"); 1464 static ConstString g_sect_name_dwarf_apple_types("__apple_types"); 1465 static ConstString g_sect_name_dwarf_apple_namespaces("__apple_namespac"); 1466 static ConstString g_sect_name_dwarf_apple_objc("__apple_objc"); 1467 static ConstString g_sect_name_eh_frame("__eh_frame"); 1468 static ConstString g_sect_name_compact_unwind("__unwind_info"); 1469 static ConstString g_sect_name_text("__text"); 1470 static ConstString g_sect_name_data("__data"); 1471 static ConstString g_sect_name_go_symtab("__gosymtab"); 1472 1473 if (section_name == g_sect_name_dwarf_debug_abbrev) 1474 return eSectionTypeDWARFDebugAbbrev; 1475 if (section_name == g_sect_name_dwarf_debug_aranges) 1476 return eSectionTypeDWARFDebugAranges; 1477 if (section_name == g_sect_name_dwarf_debug_frame) 1478 return eSectionTypeDWARFDebugFrame; 1479 if (section_name == g_sect_name_dwarf_debug_info) 1480 return eSectionTypeDWARFDebugInfo; 1481 if (section_name == g_sect_name_dwarf_debug_line) 1482 return eSectionTypeDWARFDebugLine; 1483 if (section_name == g_sect_name_dwarf_debug_loc) 1484 return eSectionTypeDWARFDebugLoc; 1485 if (section_name == g_sect_name_dwarf_debug_macinfo) 1486 return eSectionTypeDWARFDebugMacInfo; 1487 if (section_name == g_sect_name_dwarf_debug_pubnames) 1488 return eSectionTypeDWARFDebugPubNames; 1489 if (section_name == g_sect_name_dwarf_debug_pubtypes) 1490 return eSectionTypeDWARFDebugPubTypes; 1491 if (section_name == g_sect_name_dwarf_debug_ranges) 1492 return eSectionTypeDWARFDebugRanges; 1493 if (section_name == g_sect_name_dwarf_debug_str) 1494 return eSectionTypeDWARFDebugStr; 1495 if (section_name == g_sect_name_dwarf_debug_types) 1496 return eSectionTypeDWARFDebugTypes; 1497 if (section_name == g_sect_name_dwarf_apple_names) 1498 return eSectionTypeDWARFAppleNames; 1499 if (section_name == g_sect_name_dwarf_apple_types) 1500 return eSectionTypeDWARFAppleTypes; 1501 if (section_name == g_sect_name_dwarf_apple_namespaces) 1502 return eSectionTypeDWARFAppleNamespaces; 1503 if (section_name == g_sect_name_dwarf_apple_objc) 1504 return eSectionTypeDWARFAppleObjC; 1505 if (section_name == g_sect_name_objc_selrefs) 1506 return eSectionTypeDataCStringPointers; 1507 if (section_name == g_sect_name_objc_msgrefs) 1508 return eSectionTypeDataObjCMessageRefs; 1509 if (section_name == g_sect_name_eh_frame) 1510 return eSectionTypeEHFrame; 1511 if (section_name == g_sect_name_compact_unwind) 1512 return eSectionTypeCompactUnwind; 1513 if (section_name == g_sect_name_cfstring) 1514 return eSectionTypeDataObjCCFStrings; 1515 if (section_name == g_sect_name_go_symtab) 1516 return eSectionTypeGoSymtab; 1517 if (section_name == g_sect_name_objc_data || 1518 section_name == g_sect_name_objc_classrefs || 1519 section_name == g_sect_name_objc_superrefs || 1520 section_name == g_sect_name_objc_const || 1521 section_name == g_sect_name_objc_classlist) { 1522 return eSectionTypeDataPointers; 1523 } 1524 1525 switch (mach_sect_type) { 1526 // TODO: categorize sections by other flags for regular sections 1527 case S_REGULAR: 1528 if (section_name == g_sect_name_text) 1529 return eSectionTypeCode; 1530 if (section_name == g_sect_name_data) 1531 return eSectionTypeData; 1532 return eSectionTypeOther; 1533 case S_ZEROFILL: 1534 return eSectionTypeZeroFill; 1535 case S_CSTRING_LITERALS: // section with only literal C strings 1536 return eSectionTypeDataCString; 1537 case S_4BYTE_LITERALS: // section with only 4 byte literals 1538 return eSectionTypeData4; 1539 case S_8BYTE_LITERALS: // section with only 8 byte literals 1540 return eSectionTypeData8; 1541 case S_LITERAL_POINTERS: // section with only pointers to literals 1542 return eSectionTypeDataPointers; 1543 case S_NON_LAZY_SYMBOL_POINTERS: // section with only non-lazy symbol pointers 1544 return eSectionTypeDataPointers; 1545 case S_LAZY_SYMBOL_POINTERS: // section with only lazy symbol pointers 1546 return eSectionTypeDataPointers; 1547 case S_SYMBOL_STUBS: // section with only symbol stubs, byte size of stub in 1548 // the reserved2 field 1549 return eSectionTypeCode; 1550 case S_MOD_INIT_FUNC_POINTERS: // section with only function pointers for 1551 // initialization 1552 return eSectionTypeDataPointers; 1553 case S_MOD_TERM_FUNC_POINTERS: // section with only function pointers for 1554 // termination 1555 return eSectionTypeDataPointers; 1556 case S_COALESCED: 1557 return eSectionTypeOther; 1558 case S_GB_ZEROFILL: 1559 return eSectionTypeZeroFill; 1560 case S_INTERPOSING: // section with only pairs of function pointers for 1561 // interposing 1562 return eSectionTypeCode; 1563 case S_16BYTE_LITERALS: // section with only 16 byte literals 1564 return eSectionTypeData16; 1565 case S_DTRACE_DOF: 1566 return eSectionTypeDebug; 1567 case S_LAZY_DYLIB_SYMBOL_POINTERS: 1568 return eSectionTypeDataPointers; 1569 default: 1570 return eSectionTypeOther; 1571 } 1572 } 1573 1574 struct ObjectFileMachO::SegmentParsingContext { 1575 const EncryptedFileRanges EncryptedRanges; 1576 lldb_private::SectionList &UnifiedList; 1577 uint32_t NextSegmentIdx = 0; 1578 uint32_t NextSectionIdx = 0; 1579 bool FileAddressesChanged = false; 1580 1581 SegmentParsingContext(EncryptedFileRanges EncryptedRanges, 1582 lldb_private::SectionList &UnifiedList) 1583 : EncryptedRanges(std::move(EncryptedRanges)), UnifiedList(UnifiedList) {} 1584 }; 1585 1586 void ObjectFileMachO::ProcessSegmentCommand(const load_command &load_cmd_, 1587 lldb::offset_t offset, 1588 uint32_t cmd_idx, 1589 SegmentParsingContext &context) { 1590 segment_command_64 load_cmd; 1591 memcpy(&load_cmd, &load_cmd_, sizeof(load_cmd_)); 1592 1593 if (!m_data.GetU8(&offset, (uint8_t *)load_cmd.segname, 16)) 1594 return; 1595 1596 ModuleSP module_sp = GetModule(); 1597 const bool is_core = GetType() == eTypeCoreFile; 1598 const bool is_dsym = (m_header.filetype == MH_DSYM); 1599 bool add_section = true; 1600 bool add_to_unified = true; 1601 ConstString const_segname( 1602 load_cmd.segname, 1603 std::min<size_t>(strlen(load_cmd.segname), sizeof(load_cmd.segname))); 1604 1605 SectionSP unified_section_sp( 1606 context.UnifiedList.FindSectionByName(const_segname)); 1607 if (is_dsym && unified_section_sp) { 1608 if (const_segname == GetSegmentNameLINKEDIT()) { 1609 // We need to keep the __LINKEDIT segment private to this object file 1610 // only 1611 add_to_unified = false; 1612 } else { 1613 // This is the dSYM file and this section has already been created by the 1614 // object file, no need to create it. 1615 add_section = false; 1616 } 1617 } 1618 load_cmd.vmaddr = m_data.GetAddress(&offset); 1619 load_cmd.vmsize = m_data.GetAddress(&offset); 1620 load_cmd.fileoff = m_data.GetAddress(&offset); 1621 load_cmd.filesize = m_data.GetAddress(&offset); 1622 if (!m_data.GetU32(&offset, &load_cmd.maxprot, 4)) 1623 return; 1624 1625 SanitizeSegmentCommand(load_cmd, cmd_idx); 1626 1627 const uint32_t segment_permissions = GetSegmentPermissions(load_cmd); 1628 const bool segment_is_encrypted = 1629 (load_cmd.flags & SG_PROTECTED_VERSION_1) != 0; 1630 1631 // Keep a list of mach segments around in case we need to get at data that 1632 // isn't stored in the abstracted Sections. 1633 m_mach_segments.push_back(load_cmd); 1634 1635 // Use a segment ID of the segment index shifted left by 8 so they never 1636 // conflict with any of the sections. 1637 SectionSP segment_sp; 1638 if (add_section && (const_segname || is_core)) { 1639 segment_sp.reset(new Section( 1640 module_sp, // Module to which this section belongs 1641 this, // Object file to which this sections belongs 1642 ++context.NextSegmentIdx 1643 << 8, // Section ID is the 1 based segment index 1644 // shifted right by 8 bits as not to collide with any of the 256 1645 // section IDs that are possible 1646 const_segname, // Name of this section 1647 eSectionTypeContainer, // This section is a container of other 1648 // sections. 1649 load_cmd.vmaddr, // File VM address == addresses as they are 1650 // found in the object file 1651 load_cmd.vmsize, // VM size in bytes of this section 1652 load_cmd.fileoff, // Offset to the data for this section in 1653 // the file 1654 load_cmd.filesize, // Size in bytes of this section as found 1655 // in the file 1656 0, // Segments have no alignment information 1657 load_cmd.flags)); // Flags for this section 1658 1659 segment_sp->SetIsEncrypted(segment_is_encrypted); 1660 m_sections_ap->AddSection(segment_sp); 1661 segment_sp->SetPermissions(segment_permissions); 1662 if (add_to_unified) 1663 context.UnifiedList.AddSection(segment_sp); 1664 } else if (unified_section_sp) { 1665 if (is_dsym && unified_section_sp->GetFileAddress() != load_cmd.vmaddr) { 1666 // Check to see if the module was read from memory? 1667 if (module_sp->GetObjectFile()->GetHeaderAddress().IsValid()) { 1668 // We have a module that is in memory and needs to have its file 1669 // address adjusted. We need to do this because when we load a file 1670 // from memory, its addresses will be slid already, yet the addresses 1671 // in the new symbol file will still be unslid. Since everything is 1672 // stored as section offset, this shouldn't cause any problems. 1673 1674 // Make sure we've parsed the symbol table from the ObjectFile before 1675 // we go around changing its Sections. 1676 module_sp->GetObjectFile()->GetSymtab(); 1677 // eh_frame would present the same problems but we parse that on a per- 1678 // function basis as-needed so it's more difficult to remove its use of 1679 // the Sections. Realistically, the environments where this code path 1680 // will be taken will not have eh_frame sections. 1681 1682 unified_section_sp->SetFileAddress(load_cmd.vmaddr); 1683 1684 // Notify the module that the section addresses have been changed once 1685 // we're done so any file-address caches can be updated. 1686 context.FileAddressesChanged = true; 1687 } 1688 } 1689 m_sections_ap->AddSection(unified_section_sp); 1690 } 1691 1692 struct section_64 sect64; 1693 ::memset(§64, 0, sizeof(sect64)); 1694 // Push a section into our mach sections for the section at index zero 1695 // (NO_SECT) if we don't have any mach sections yet... 1696 if (m_mach_sections.empty()) 1697 m_mach_sections.push_back(sect64); 1698 uint32_t segment_sect_idx; 1699 const lldb::user_id_t first_segment_sectID = context.NextSectionIdx + 1; 1700 1701 const uint32_t num_u32s = load_cmd.cmd == LC_SEGMENT ? 7 : 8; 1702 for (segment_sect_idx = 0; segment_sect_idx < load_cmd.nsects; 1703 ++segment_sect_idx) { 1704 if (m_data.GetU8(&offset, (uint8_t *)sect64.sectname, 1705 sizeof(sect64.sectname)) == NULL) 1706 break; 1707 if (m_data.GetU8(&offset, (uint8_t *)sect64.segname, 1708 sizeof(sect64.segname)) == NULL) 1709 break; 1710 sect64.addr = m_data.GetAddress(&offset); 1711 sect64.size = m_data.GetAddress(&offset); 1712 1713 if (m_data.GetU32(&offset, §64.offset, num_u32s) == NULL) 1714 break; 1715 1716 // Keep a list of mach sections around in case we need to get at data that 1717 // isn't stored in the abstracted Sections. 1718 m_mach_sections.push_back(sect64); 1719 1720 if (add_section) { 1721 ConstString section_name( 1722 sect64.sectname, 1723 std::min<size_t>(strlen(sect64.sectname), sizeof(sect64.sectname))); 1724 if (!const_segname) { 1725 // We have a segment with no name so we need to conjure up segments 1726 // that correspond to the section's segname if there isn't already such 1727 // a section. If there is such a section, we resize the section so that 1728 // it spans all sections. We also mark these sections as fake so 1729 // address matches don't hit if they land in the gaps between the child 1730 // sections. 1731 const_segname.SetTrimmedCStringWithLength(sect64.segname, 1732 sizeof(sect64.segname)); 1733 segment_sp = context.UnifiedList.FindSectionByName(const_segname); 1734 if (segment_sp.get()) { 1735 Section *segment = segment_sp.get(); 1736 // Grow the section size as needed. 1737 const lldb::addr_t sect64_min_addr = sect64.addr; 1738 const lldb::addr_t sect64_max_addr = sect64_min_addr + sect64.size; 1739 const lldb::addr_t curr_seg_byte_size = segment->GetByteSize(); 1740 const lldb::addr_t curr_seg_min_addr = segment->GetFileAddress(); 1741 const lldb::addr_t curr_seg_max_addr = 1742 curr_seg_min_addr + curr_seg_byte_size; 1743 if (sect64_min_addr >= curr_seg_min_addr) { 1744 const lldb::addr_t new_seg_byte_size = 1745 sect64_max_addr - curr_seg_min_addr; 1746 // Only grow the section size if needed 1747 if (new_seg_byte_size > curr_seg_byte_size) 1748 segment->SetByteSize(new_seg_byte_size); 1749 } else { 1750 // We need to change the base address of the segment and adjust the 1751 // child section offsets for all existing children. 1752 const lldb::addr_t slide_amount = 1753 sect64_min_addr - curr_seg_min_addr; 1754 segment->Slide(slide_amount, false); 1755 segment->GetChildren().Slide(-slide_amount, false); 1756 segment->SetByteSize(curr_seg_max_addr - sect64_min_addr); 1757 } 1758 1759 // Grow the section size as needed. 1760 if (sect64.offset) { 1761 const lldb::addr_t segment_min_file_offset = 1762 segment->GetFileOffset(); 1763 const lldb::addr_t segment_max_file_offset = 1764 segment_min_file_offset + segment->GetFileSize(); 1765 1766 const lldb::addr_t section_min_file_offset = sect64.offset; 1767 const lldb::addr_t section_max_file_offset = 1768 section_min_file_offset + sect64.size; 1769 const lldb::addr_t new_file_offset = 1770 std::min(section_min_file_offset, segment_min_file_offset); 1771 const lldb::addr_t new_file_size = 1772 std::max(section_max_file_offset, segment_max_file_offset) - 1773 new_file_offset; 1774 segment->SetFileOffset(new_file_offset); 1775 segment->SetFileSize(new_file_size); 1776 } 1777 } else { 1778 // Create a fake section for the section's named segment 1779 segment_sp.reset(new Section( 1780 segment_sp, // Parent section 1781 module_sp, // Module to which this section belongs 1782 this, // Object file to which this section belongs 1783 ++context.NextSegmentIdx 1784 << 8, // Section ID is the 1 based segment index 1785 // shifted right by 8 bits as not to 1786 // collide with any of the 256 section IDs 1787 // that are possible 1788 const_segname, // Name of this section 1789 eSectionTypeContainer, // This section is a container of 1790 // other sections. 1791 sect64.addr, // File VM address == addresses as they are 1792 // found in the object file 1793 sect64.size, // VM size in bytes of this section 1794 sect64.offset, // Offset to the data for this section in 1795 // the file 1796 sect64.offset ? sect64.size : 0, // Size in bytes of 1797 // this section as 1798 // found in the file 1799 sect64.align, 1800 load_cmd.flags)); // Flags for this section 1801 segment_sp->SetIsFake(true); 1802 segment_sp->SetPermissions(segment_permissions); 1803 m_sections_ap->AddSection(segment_sp); 1804 if (add_to_unified) 1805 context.UnifiedList.AddSection(segment_sp); 1806 segment_sp->SetIsEncrypted(segment_is_encrypted); 1807 } 1808 } 1809 assert(segment_sp.get()); 1810 1811 lldb::SectionType sect_type = GetSectionType(sect64.flags, section_name); 1812 1813 SectionSP section_sp(new Section( 1814 segment_sp, module_sp, this, ++context.NextSectionIdx, section_name, 1815 sect_type, sect64.addr - segment_sp->GetFileAddress(), sect64.size, 1816 sect64.offset, sect64.offset == 0 ? 0 : sect64.size, sect64.align, 1817 sect64.flags)); 1818 // Set the section to be encrypted to match the segment 1819 1820 bool section_is_encrypted = false; 1821 if (!segment_is_encrypted && load_cmd.filesize != 0) 1822 section_is_encrypted = context.EncryptedRanges.FindEntryThatContains( 1823 sect64.offset) != NULL; 1824 1825 section_sp->SetIsEncrypted(segment_is_encrypted || section_is_encrypted); 1826 section_sp->SetPermissions(segment_permissions); 1827 segment_sp->GetChildren().AddSection(section_sp); 1828 1829 if (segment_sp->IsFake()) { 1830 segment_sp.reset(); 1831 const_segname.Clear(); 1832 } 1833 } 1834 } 1835 if (segment_sp && is_dsym) { 1836 if (first_segment_sectID <= context.NextSectionIdx) { 1837 lldb::user_id_t sect_uid; 1838 for (sect_uid = first_segment_sectID; sect_uid <= context.NextSectionIdx; 1839 ++sect_uid) { 1840 SectionSP curr_section_sp( 1841 segment_sp->GetChildren().FindSectionByID(sect_uid)); 1842 SectionSP next_section_sp; 1843 if (sect_uid + 1 <= context.NextSectionIdx) 1844 next_section_sp = 1845 segment_sp->GetChildren().FindSectionByID(sect_uid + 1); 1846 1847 if (curr_section_sp.get()) { 1848 if (curr_section_sp->GetByteSize() == 0) { 1849 if (next_section_sp.get() != NULL) 1850 curr_section_sp->SetByteSize(next_section_sp->GetFileAddress() - 1851 curr_section_sp->GetFileAddress()); 1852 else 1853 curr_section_sp->SetByteSize(load_cmd.vmsize); 1854 } 1855 } 1856 } 1857 } 1858 } 1859 } 1860 1861 void ObjectFileMachO::ProcessDysymtabCommand(const load_command &load_cmd, 1862 lldb::offset_t offset) { 1863 m_dysymtab.cmd = load_cmd.cmd; 1864 m_dysymtab.cmdsize = load_cmd.cmdsize; 1865 m_data.GetU32(&offset, &m_dysymtab.ilocalsym, 1866 (sizeof(m_dysymtab) / sizeof(uint32_t)) - 2); 1867 } 1868 1869 void ObjectFileMachO::CreateSections(SectionList &unified_section_list) { 1870 if (m_sections_ap) 1871 return; 1872 1873 m_sections_ap.reset(new SectionList()); 1874 1875 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 1876 // bool dump_sections = false; 1877 ModuleSP module_sp(GetModule()); 1878 1879 offset = MachHeaderSizeFromMagic(m_header.magic); 1880 1881 SegmentParsingContext context(GetEncryptedFileRanges(), unified_section_list); 1882 struct load_command load_cmd; 1883 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 1884 const lldb::offset_t load_cmd_offset = offset; 1885 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 1886 break; 1887 1888 if (load_cmd.cmd == LC_SEGMENT || load_cmd.cmd == LC_SEGMENT_64) 1889 ProcessSegmentCommand(load_cmd, offset, i, context); 1890 else if (load_cmd.cmd == LC_DYSYMTAB) 1891 ProcessDysymtabCommand(load_cmd, offset); 1892 1893 offset = load_cmd_offset + load_cmd.cmdsize; 1894 } 1895 1896 if (context.FileAddressesChanged && module_sp) 1897 module_sp->SectionFileAddressesChanged(); 1898 } 1899 1900 class MachSymtabSectionInfo { 1901 public: 1902 MachSymtabSectionInfo(SectionList *section_list) 1903 : m_section_list(section_list), m_section_infos() { 1904 // Get the number of sections down to a depth of 1 to include all segments 1905 // and their sections, but no other sections that may be added for debug 1906 // map or 1907 m_section_infos.resize(section_list->GetNumSections(1)); 1908 } 1909 1910 SectionSP GetSection(uint8_t n_sect, addr_t file_addr) { 1911 if (n_sect == 0) 1912 return SectionSP(); 1913 if (n_sect < m_section_infos.size()) { 1914 if (!m_section_infos[n_sect].section_sp) { 1915 SectionSP section_sp(m_section_list->FindSectionByID(n_sect)); 1916 m_section_infos[n_sect].section_sp = section_sp; 1917 if (section_sp) { 1918 m_section_infos[n_sect].vm_range.SetBaseAddress( 1919 section_sp->GetFileAddress()); 1920 m_section_infos[n_sect].vm_range.SetByteSize( 1921 section_sp->GetByteSize()); 1922 } else { 1923 Host::SystemLog(Host::eSystemLogError, 1924 "error: unable to find section for section %u\n", 1925 n_sect); 1926 } 1927 } 1928 if (m_section_infos[n_sect].vm_range.Contains(file_addr)) { 1929 // Symbol is in section. 1930 return m_section_infos[n_sect].section_sp; 1931 } else if (m_section_infos[n_sect].vm_range.GetByteSize() == 0 && 1932 m_section_infos[n_sect].vm_range.GetBaseAddress() == 1933 file_addr) { 1934 // Symbol is in section with zero size, but has the same start address 1935 // as the section. This can happen with linker symbols (symbols that 1936 // start with the letter 'l' or 'L'. 1937 return m_section_infos[n_sect].section_sp; 1938 } 1939 } 1940 return m_section_list->FindSectionContainingFileAddress(file_addr); 1941 } 1942 1943 protected: 1944 struct SectionInfo { 1945 SectionInfo() : vm_range(), section_sp() {} 1946 1947 VMRange vm_range; 1948 SectionSP section_sp; 1949 }; 1950 SectionList *m_section_list; 1951 std::vector<SectionInfo> m_section_infos; 1952 }; 1953 1954 struct TrieEntry { 1955 TrieEntry() 1956 : name(), address(LLDB_INVALID_ADDRESS), flags(0), other(0), 1957 import_name() {} 1958 1959 void Clear() { 1960 name.Clear(); 1961 address = LLDB_INVALID_ADDRESS; 1962 flags = 0; 1963 other = 0; 1964 import_name.Clear(); 1965 } 1966 1967 void Dump() const { 1968 printf("0x%16.16llx 0x%16.16llx 0x%16.16llx \"%s\"", 1969 static_cast<unsigned long long>(address), 1970 static_cast<unsigned long long>(flags), 1971 static_cast<unsigned long long>(other), name.GetCString()); 1972 if (import_name) 1973 printf(" -> \"%s\"\n", import_name.GetCString()); 1974 else 1975 printf("\n"); 1976 } 1977 ConstString name; 1978 uint64_t address; 1979 uint64_t flags; 1980 uint64_t other; 1981 ConstString import_name; 1982 }; 1983 1984 struct TrieEntryWithOffset { 1985 lldb::offset_t nodeOffset; 1986 TrieEntry entry; 1987 1988 TrieEntryWithOffset(lldb::offset_t offset) : nodeOffset(offset), entry() {} 1989 1990 void Dump(uint32_t idx) const { 1991 printf("[%3u] 0x%16.16llx: ", idx, 1992 static_cast<unsigned long long>(nodeOffset)); 1993 entry.Dump(); 1994 } 1995 1996 bool operator<(const TrieEntryWithOffset &other) const { 1997 return (nodeOffset < other.nodeOffset); 1998 } 1999 }; 2000 2001 static bool ParseTrieEntries(DataExtractor &data, lldb::offset_t offset, 2002 const bool is_arm, 2003 std::vector<llvm::StringRef> &nameSlices, 2004 std::set<lldb::addr_t> &resolver_addresses, 2005 std::vector<TrieEntryWithOffset> &output) { 2006 if (!data.ValidOffset(offset)) 2007 return true; 2008 2009 const uint64_t terminalSize = data.GetULEB128(&offset); 2010 lldb::offset_t children_offset = offset + terminalSize; 2011 if (terminalSize != 0) { 2012 TrieEntryWithOffset e(offset); 2013 e.entry.flags = data.GetULEB128(&offset); 2014 const char *import_name = NULL; 2015 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_REEXPORT) { 2016 e.entry.address = 0; 2017 e.entry.other = data.GetULEB128(&offset); // dylib ordinal 2018 import_name = data.GetCStr(&offset); 2019 } else { 2020 e.entry.address = data.GetULEB128(&offset); 2021 if (e.entry.flags & EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER) { 2022 e.entry.other = data.GetULEB128(&offset); 2023 uint64_t resolver_addr = e.entry.other; 2024 if (is_arm) 2025 resolver_addr &= THUMB_ADDRESS_BIT_MASK; 2026 resolver_addresses.insert(resolver_addr); 2027 } else 2028 e.entry.other = 0; 2029 } 2030 // Only add symbols that are reexport symbols with a valid import name 2031 if (EXPORT_SYMBOL_FLAGS_REEXPORT & e.entry.flags && import_name && 2032 import_name[0]) { 2033 std::string name; 2034 if (!nameSlices.empty()) { 2035 for (auto name_slice : nameSlices) 2036 name.append(name_slice.data(), name_slice.size()); 2037 } 2038 if (name.size() > 1) { 2039 // Skip the leading '_' 2040 e.entry.name.SetCStringWithLength(name.c_str() + 1, name.size() - 1); 2041 } 2042 if (import_name) { 2043 // Skip the leading '_' 2044 e.entry.import_name.SetCString(import_name + 1); 2045 } 2046 output.push_back(e); 2047 } 2048 } 2049 2050 const uint8_t childrenCount = data.GetU8(&children_offset); 2051 for (uint8_t i = 0; i < childrenCount; ++i) { 2052 const char *cstr = data.GetCStr(&children_offset); 2053 if (cstr) 2054 nameSlices.push_back(llvm::StringRef(cstr)); 2055 else 2056 return false; // Corrupt data 2057 lldb::offset_t childNodeOffset = data.GetULEB128(&children_offset); 2058 if (childNodeOffset) { 2059 if (!ParseTrieEntries(data, childNodeOffset, is_arm, nameSlices, 2060 resolver_addresses, output)) { 2061 return false; 2062 } 2063 } 2064 nameSlices.pop_back(); 2065 } 2066 return true; 2067 } 2068 2069 // Read the UUID out of a dyld_shared_cache file on-disk. 2070 UUID ObjectFileMachO::GetSharedCacheUUID(FileSpec dyld_shared_cache, 2071 const ByteOrder byte_order, 2072 const uint32_t addr_byte_size) { 2073 UUID dsc_uuid; 2074 DataBufferSP DscData = MapFileData( 2075 dyld_shared_cache, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2076 if (!DscData) 2077 return dsc_uuid; 2078 DataExtractor dsc_header_data(DscData, byte_order, addr_byte_size); 2079 2080 char version_str[7]; 2081 lldb::offset_t offset = 0; 2082 memcpy(version_str, dsc_header_data.GetData(&offset, 6), 6); 2083 version_str[6] = '\0'; 2084 if (strcmp(version_str, "dyld_v") == 0) { 2085 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, uuid); 2086 uint8_t uuid_bytes[sizeof(uuid_t)]; 2087 memcpy(uuid_bytes, dsc_header_data.GetData(&offset, sizeof(uuid_t)), 2088 sizeof(uuid_t)); 2089 dsc_uuid.SetBytes(uuid_bytes); 2090 } 2091 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2092 if (log && dsc_uuid.IsValid()) { 2093 log->Printf("Shared cache %s has UUID %s", dyld_shared_cache.GetPath().c_str(), 2094 dsc_uuid.GetAsString().c_str()); 2095 } 2096 return dsc_uuid; 2097 } 2098 2099 size_t ObjectFileMachO::ParseSymtab() { 2100 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION); 2101 Timer scoped_timer(func_cat, "ObjectFileMachO::ParseSymtab () module = %s", 2102 m_file.GetFilename().AsCString("")); 2103 ModuleSP module_sp(GetModule()); 2104 if (!module_sp) 2105 return 0; 2106 2107 struct symtab_command symtab_load_command = {0, 0, 0, 0, 0, 0}; 2108 struct linkedit_data_command function_starts_load_command = {0, 0, 0, 0}; 2109 struct dyld_info_command dyld_info = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}; 2110 typedef AddressDataArray<lldb::addr_t, bool, 100> FunctionStarts; 2111 FunctionStarts function_starts; 2112 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 2113 uint32_t i; 2114 FileSpecList dylib_files; 2115 Log *log(lldb_private::GetLogIfAllCategoriesSet(LIBLLDB_LOG_SYMBOLS)); 2116 static const llvm::StringRef g_objc_v2_prefix_class("_OBJC_CLASS_$_"); 2117 static const llvm::StringRef g_objc_v2_prefix_metaclass("_OBJC_METACLASS_$_"); 2118 static const llvm::StringRef g_objc_v2_prefix_ivar("_OBJC_IVAR_$_"); 2119 2120 for (i = 0; i < m_header.ncmds; ++i) { 2121 const lldb::offset_t cmd_offset = offset; 2122 // Read in the load command and load command size 2123 struct load_command lc; 2124 if (m_data.GetU32(&offset, &lc, 2) == NULL) 2125 break; 2126 // Watch for the symbol table load command 2127 switch (lc.cmd) { 2128 case LC_SYMTAB: 2129 symtab_load_command.cmd = lc.cmd; 2130 symtab_load_command.cmdsize = lc.cmdsize; 2131 // Read in the rest of the symtab load command 2132 if (m_data.GetU32(&offset, &symtab_load_command.symoff, 4) == 2133 0) // fill in symoff, nsyms, stroff, strsize fields 2134 return 0; 2135 if (symtab_load_command.symoff == 0) { 2136 if (log) 2137 module_sp->LogMessage(log, "LC_SYMTAB.symoff == 0"); 2138 return 0; 2139 } 2140 2141 if (symtab_load_command.stroff == 0) { 2142 if (log) 2143 module_sp->LogMessage(log, "LC_SYMTAB.stroff == 0"); 2144 return 0; 2145 } 2146 2147 if (symtab_load_command.nsyms == 0) { 2148 if (log) 2149 module_sp->LogMessage(log, "LC_SYMTAB.nsyms == 0"); 2150 return 0; 2151 } 2152 2153 if (symtab_load_command.strsize == 0) { 2154 if (log) 2155 module_sp->LogMessage(log, "LC_SYMTAB.strsize == 0"); 2156 return 0; 2157 } 2158 break; 2159 2160 case LC_DYLD_INFO: 2161 case LC_DYLD_INFO_ONLY: 2162 if (m_data.GetU32(&offset, &dyld_info.rebase_off, 10)) { 2163 dyld_info.cmd = lc.cmd; 2164 dyld_info.cmdsize = lc.cmdsize; 2165 } else { 2166 memset(&dyld_info, 0, sizeof(dyld_info)); 2167 } 2168 break; 2169 2170 case LC_LOAD_DYLIB: 2171 case LC_LOAD_WEAK_DYLIB: 2172 case LC_REEXPORT_DYLIB: 2173 case LC_LOADFVMLIB: 2174 case LC_LOAD_UPWARD_DYLIB: { 2175 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 2176 const char *path = m_data.PeekCStr(name_offset); 2177 if (path) { 2178 FileSpec file_spec(path, false); 2179 // Strip the path if there is @rpath, @executable, etc so we just use 2180 // the basename 2181 if (path[0] == '@') 2182 file_spec.GetDirectory().Clear(); 2183 2184 if (lc.cmd == LC_REEXPORT_DYLIB) { 2185 m_reexported_dylibs.AppendIfUnique(file_spec); 2186 } 2187 2188 dylib_files.Append(file_spec); 2189 } 2190 } break; 2191 2192 case LC_FUNCTION_STARTS: 2193 function_starts_load_command.cmd = lc.cmd; 2194 function_starts_load_command.cmdsize = lc.cmdsize; 2195 if (m_data.GetU32(&offset, &function_starts_load_command.dataoff, 2) == 2196 NULL) // fill in symoff, nsyms, stroff, strsize fields 2197 memset(&function_starts_load_command, 0, 2198 sizeof(function_starts_load_command)); 2199 break; 2200 2201 default: 2202 break; 2203 } 2204 offset = cmd_offset + lc.cmdsize; 2205 } 2206 2207 if (symtab_load_command.cmd) { 2208 Symtab *symtab = m_symtab_ap.get(); 2209 SectionList *section_list = GetSectionList(); 2210 if (section_list == NULL) 2211 return 0; 2212 2213 const uint32_t addr_byte_size = m_data.GetAddressByteSize(); 2214 const ByteOrder byte_order = m_data.GetByteOrder(); 2215 bool bit_width_32 = addr_byte_size == 4; 2216 const size_t nlist_byte_size = 2217 bit_width_32 ? sizeof(struct nlist) : sizeof(struct nlist_64); 2218 2219 DataExtractor nlist_data(NULL, 0, byte_order, addr_byte_size); 2220 DataExtractor strtab_data(NULL, 0, byte_order, addr_byte_size); 2221 DataExtractor function_starts_data(NULL, 0, byte_order, addr_byte_size); 2222 DataExtractor indirect_symbol_index_data(NULL, 0, byte_order, 2223 addr_byte_size); 2224 DataExtractor dyld_trie_data(NULL, 0, byte_order, addr_byte_size); 2225 2226 const addr_t nlist_data_byte_size = 2227 symtab_load_command.nsyms * nlist_byte_size; 2228 const addr_t strtab_data_byte_size = symtab_load_command.strsize; 2229 addr_t strtab_addr = LLDB_INVALID_ADDRESS; 2230 2231 ProcessSP process_sp(m_process_wp.lock()); 2232 Process *process = process_sp.get(); 2233 2234 uint32_t memory_module_load_level = eMemoryModuleLoadLevelComplete; 2235 2236 if (process && m_header.filetype != llvm::MachO::MH_OBJECT) { 2237 Target &target = process->GetTarget(); 2238 2239 memory_module_load_level = target.GetMemoryModuleLoadLevel(); 2240 2241 SectionSP linkedit_section_sp( 2242 section_list->FindSectionByName(GetSegmentNameLINKEDIT())); 2243 // Reading mach file from memory in a process or core file... 2244 2245 if (linkedit_section_sp) { 2246 addr_t linkedit_load_addr = 2247 linkedit_section_sp->GetLoadBaseAddress(&target); 2248 if (linkedit_load_addr == LLDB_INVALID_ADDRESS) { 2249 // We might be trying to access the symbol table before the 2250 // __LINKEDIT's load address has been set in the target. We can't 2251 // fail to read the symbol table, so calculate the right address 2252 // manually 2253 linkedit_load_addr = CalculateSectionLoadAddressForMemoryImage( 2254 m_memory_addr, GetMachHeaderSection(), linkedit_section_sp.get()); 2255 } 2256 2257 const addr_t linkedit_file_offset = 2258 linkedit_section_sp->GetFileOffset(); 2259 const addr_t symoff_addr = linkedit_load_addr + 2260 symtab_load_command.symoff - 2261 linkedit_file_offset; 2262 strtab_addr = linkedit_load_addr + symtab_load_command.stroff - 2263 linkedit_file_offset; 2264 2265 bool data_was_read = false; 2266 2267 #if defined(__APPLE__) && \ 2268 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2269 if (m_header.flags & 0x80000000u && 2270 process->GetAddressByteSize() == sizeof(void *)) { 2271 // This mach-o memory file is in the dyld shared cache. If this 2272 // program is not remote and this is iOS, then this process will 2273 // share the same shared cache as the process we are debugging and we 2274 // can read the entire __LINKEDIT from the address space in this 2275 // process. This is a needed optimization that is used for local iOS 2276 // debugging only since all shared libraries in the shared cache do 2277 // not have corresponding files that exist in the file system of the 2278 // device. They have been combined into a single file. This means we 2279 // always have to load these files from memory. All of the symbol and 2280 // string tables from all of the __LINKEDIT sections from the shared 2281 // libraries in the shared cache have been merged into a single large 2282 // symbol and string table. Reading all of this symbol and string 2283 // table data across can slow down debug launch times, so we optimize 2284 // this by reading the memory for the __LINKEDIT section from this 2285 // process. 2286 2287 UUID lldb_shared_cache; 2288 addr_t lldb_shared_cache_addr; 2289 GetLLDBSharedCacheUUID (lldb_shared_cache_addr, lldb_shared_cache); 2290 UUID process_shared_cache; 2291 addr_t process_shared_cache_addr; 2292 GetProcessSharedCacheUUID(process, process_shared_cache_addr, process_shared_cache); 2293 bool use_lldb_cache = true; 2294 if (lldb_shared_cache.IsValid() && process_shared_cache.IsValid() && 2295 (lldb_shared_cache != process_shared_cache 2296 || process_shared_cache_addr != lldb_shared_cache_addr)) { 2297 use_lldb_cache = false; 2298 } 2299 2300 PlatformSP platform_sp(target.GetPlatform()); 2301 if (platform_sp && platform_sp->IsHost() && use_lldb_cache) { 2302 data_was_read = true; 2303 nlist_data.SetData((void *)symoff_addr, nlist_data_byte_size, 2304 eByteOrderLittle); 2305 strtab_data.SetData((void *)strtab_addr, strtab_data_byte_size, 2306 eByteOrderLittle); 2307 if (function_starts_load_command.cmd) { 2308 const addr_t func_start_addr = 2309 linkedit_load_addr + function_starts_load_command.dataoff - 2310 linkedit_file_offset; 2311 function_starts_data.SetData( 2312 (void *)func_start_addr, 2313 function_starts_load_command.datasize, eByteOrderLittle); 2314 } 2315 } 2316 } 2317 #endif 2318 2319 if (!data_was_read) { 2320 // Always load dyld - the dynamic linker - from memory if we didn't 2321 // find a binary anywhere else. lldb will not register 2322 // dylib/framework/bundle loads/unloads if we don't have the dyld 2323 // symbols, we force dyld to load from memory despite the user's 2324 // target.memory-module-load-level setting. 2325 if (memory_module_load_level == eMemoryModuleLoadLevelComplete || 2326 m_header.filetype == llvm::MachO::MH_DYLINKER) { 2327 DataBufferSP nlist_data_sp( 2328 ReadMemory(process_sp, symoff_addr, nlist_data_byte_size)); 2329 if (nlist_data_sp) 2330 nlist_data.SetData(nlist_data_sp, 0, 2331 nlist_data_sp->GetByteSize()); 2332 // Load strings individually from memory when loading from memory 2333 // since shared cache string tables contain strings for all symbols 2334 // from all shared cached libraries DataBufferSP strtab_data_sp 2335 // (ReadMemory (process_sp, strtab_addr, 2336 // strtab_data_byte_size)); 2337 // if (strtab_data_sp) 2338 // strtab_data.SetData (strtab_data_sp, 0, 2339 // strtab_data_sp->GetByteSize()); 2340 if (m_dysymtab.nindirectsyms != 0) { 2341 const addr_t indirect_syms_addr = linkedit_load_addr + 2342 m_dysymtab.indirectsymoff - 2343 linkedit_file_offset; 2344 DataBufferSP indirect_syms_data_sp( 2345 ReadMemory(process_sp, indirect_syms_addr, 2346 m_dysymtab.nindirectsyms * 4)); 2347 if (indirect_syms_data_sp) 2348 indirect_symbol_index_data.SetData( 2349 indirect_syms_data_sp, 0, 2350 indirect_syms_data_sp->GetByteSize()); 2351 } 2352 } else if (memory_module_load_level >= 2353 eMemoryModuleLoadLevelPartial) { 2354 if (function_starts_load_command.cmd) { 2355 const addr_t func_start_addr = 2356 linkedit_load_addr + function_starts_load_command.dataoff - 2357 linkedit_file_offset; 2358 DataBufferSP func_start_data_sp( 2359 ReadMemory(process_sp, func_start_addr, 2360 function_starts_load_command.datasize)); 2361 if (func_start_data_sp) 2362 function_starts_data.SetData(func_start_data_sp, 0, 2363 func_start_data_sp->GetByteSize()); 2364 } 2365 } 2366 } 2367 } 2368 } else { 2369 nlist_data.SetData(m_data, symtab_load_command.symoff, 2370 nlist_data_byte_size); 2371 strtab_data.SetData(m_data, symtab_load_command.stroff, 2372 strtab_data_byte_size); 2373 2374 if (dyld_info.export_size > 0) { 2375 dyld_trie_data.SetData(m_data, dyld_info.export_off, 2376 dyld_info.export_size); 2377 } 2378 2379 if (m_dysymtab.nindirectsyms != 0) { 2380 indirect_symbol_index_data.SetData(m_data, m_dysymtab.indirectsymoff, 2381 m_dysymtab.nindirectsyms * 4); 2382 } 2383 if (function_starts_load_command.cmd) { 2384 function_starts_data.SetData(m_data, 2385 function_starts_load_command.dataoff, 2386 function_starts_load_command.datasize); 2387 } 2388 } 2389 2390 if (nlist_data.GetByteSize() == 0 && 2391 memory_module_load_level == eMemoryModuleLoadLevelComplete) { 2392 if (log) 2393 module_sp->LogMessage(log, "failed to read nlist data"); 2394 return 0; 2395 } 2396 2397 const bool have_strtab_data = strtab_data.GetByteSize() > 0; 2398 if (!have_strtab_data) { 2399 if (process) { 2400 if (strtab_addr == LLDB_INVALID_ADDRESS) { 2401 if (log) 2402 module_sp->LogMessage(log, "failed to locate the strtab in memory"); 2403 return 0; 2404 } 2405 } else { 2406 if (log) 2407 module_sp->LogMessage(log, "failed to read strtab data"); 2408 return 0; 2409 } 2410 } 2411 2412 const ConstString &g_segment_name_TEXT = GetSegmentNameTEXT(); 2413 const ConstString &g_segment_name_DATA = GetSegmentNameDATA(); 2414 const ConstString &g_segment_name_DATA_DIRTY = GetSegmentNameDATA_DIRTY(); 2415 const ConstString &g_segment_name_DATA_CONST = GetSegmentNameDATA_CONST(); 2416 const ConstString &g_segment_name_OBJC = GetSegmentNameOBJC(); 2417 const ConstString &g_section_name_eh_frame = GetSectionNameEHFrame(); 2418 SectionSP text_section_sp( 2419 section_list->FindSectionByName(g_segment_name_TEXT)); 2420 SectionSP data_section_sp( 2421 section_list->FindSectionByName(g_segment_name_DATA)); 2422 SectionSP data_dirty_section_sp( 2423 section_list->FindSectionByName(g_segment_name_DATA_DIRTY)); 2424 SectionSP data_const_section_sp( 2425 section_list->FindSectionByName(g_segment_name_DATA_CONST)); 2426 SectionSP objc_section_sp( 2427 section_list->FindSectionByName(g_segment_name_OBJC)); 2428 SectionSP eh_frame_section_sp; 2429 if (text_section_sp.get()) 2430 eh_frame_section_sp = text_section_sp->GetChildren().FindSectionByName( 2431 g_section_name_eh_frame); 2432 else 2433 eh_frame_section_sp = 2434 section_list->FindSectionByName(g_section_name_eh_frame); 2435 2436 const bool is_arm = (m_header.cputype == llvm::MachO::CPU_TYPE_ARM); 2437 2438 // lldb works best if it knows the start address of all functions in a 2439 // module. Linker symbols or debug info are normally the best source of 2440 // information for start addr / size but they may be stripped in a released 2441 // binary. Two additional sources of information exist in Mach-O binaries: 2442 // LC_FUNCTION_STARTS - a list of ULEB128 encoded offsets of each 2443 // function's start address in the 2444 // binary, relative to the text section. 2445 // eh_frame - the eh_frame FDEs have the start addr & size of 2446 // each function 2447 // LC_FUNCTION_STARTS is the fastest source to read in, and is present on 2448 // all modern binaries. 2449 // Binaries built to run on older releases may need to use eh_frame 2450 // information. 2451 2452 if (text_section_sp && function_starts_data.GetByteSize()) { 2453 FunctionStarts::Entry function_start_entry; 2454 function_start_entry.data = false; 2455 lldb::offset_t function_start_offset = 0; 2456 function_start_entry.addr = text_section_sp->GetFileAddress(); 2457 uint64_t delta; 2458 while ((delta = function_starts_data.GetULEB128(&function_start_offset)) > 2459 0) { 2460 // Now append the current entry 2461 function_start_entry.addr += delta; 2462 function_starts.Append(function_start_entry); 2463 } 2464 } else { 2465 // If m_type is eTypeDebugInfo, then this is a dSYM - it will have the 2466 // load command claiming an eh_frame but it doesn't actually have the 2467 // eh_frame content. And if we have a dSYM, we don't need to do any of 2468 // this fill-in-the-missing-symbols works anyway - the debug info should 2469 // give us all the functions in the module. 2470 if (text_section_sp.get() && eh_frame_section_sp.get() && 2471 m_type != eTypeDebugInfo) { 2472 DWARFCallFrameInfo eh_frame(*this, eh_frame_section_sp, 2473 DWARFCallFrameInfo::EH); 2474 DWARFCallFrameInfo::FunctionAddressAndSizeVector functions; 2475 eh_frame.GetFunctionAddressAndSizeVector(functions); 2476 addr_t text_base_addr = text_section_sp->GetFileAddress(); 2477 size_t count = functions.GetSize(); 2478 for (size_t i = 0; i < count; ++i) { 2479 const DWARFCallFrameInfo::FunctionAddressAndSizeVector::Entry *func = 2480 functions.GetEntryAtIndex(i); 2481 if (func) { 2482 FunctionStarts::Entry function_start_entry; 2483 function_start_entry.addr = func->base - text_base_addr; 2484 function_starts.Append(function_start_entry); 2485 } 2486 } 2487 } 2488 } 2489 2490 const size_t function_starts_count = function_starts.GetSize(); 2491 2492 // For user process binaries (executables, dylibs, frameworks, bundles), if 2493 // we don't have LC_FUNCTION_STARTS/eh_frame section in this binary, we're 2494 // going to assume the binary has been stripped. Don't allow assembly 2495 // language instruction emulation because we don't know proper function 2496 // start boundaries. 2497 // 2498 // For all other types of binaries (kernels, stand-alone bare board 2499 // binaries, kexts), they may not have LC_FUNCTION_STARTS / eh_frame 2500 // sections - we should not make any assumptions about them based on that. 2501 if (function_starts_count == 0 && CalculateStrata() == eStrataUser) { 2502 m_allow_assembly_emulation_unwind_plans = false; 2503 Log *unwind_or_symbol_log(lldb_private::GetLogIfAnyCategoriesSet( 2504 LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_UNWIND)); 2505 2506 if (unwind_or_symbol_log) 2507 module_sp->LogMessage( 2508 unwind_or_symbol_log, 2509 "no LC_FUNCTION_STARTS, will not allow assembly profiled unwinds"); 2510 } 2511 2512 const user_id_t TEXT_eh_frame_sectID = 2513 eh_frame_section_sp.get() ? eh_frame_section_sp->GetID() 2514 : static_cast<user_id_t>(NO_SECT); 2515 2516 lldb::offset_t nlist_data_offset = 0; 2517 2518 uint32_t N_SO_index = UINT32_MAX; 2519 2520 MachSymtabSectionInfo section_info(section_list); 2521 std::vector<uint32_t> N_FUN_indexes; 2522 std::vector<uint32_t> N_NSYM_indexes; 2523 std::vector<uint32_t> N_INCL_indexes; 2524 std::vector<uint32_t> N_BRAC_indexes; 2525 std::vector<uint32_t> N_COMM_indexes; 2526 typedef std::multimap<uint64_t, uint32_t> ValueToSymbolIndexMap; 2527 typedef std::map<uint32_t, uint32_t> NListIndexToSymbolIndexMap; 2528 typedef std::map<const char *, uint32_t> ConstNameToSymbolIndexMap; 2529 ValueToSymbolIndexMap N_FUN_addr_to_sym_idx; 2530 ValueToSymbolIndexMap N_STSYM_addr_to_sym_idx; 2531 ConstNameToSymbolIndexMap N_GSYM_name_to_sym_idx; 2532 // Any symbols that get merged into another will get an entry in this map 2533 // so we know 2534 NListIndexToSymbolIndexMap m_nlist_idx_to_sym_idx; 2535 uint32_t nlist_idx = 0; 2536 Symbol *symbol_ptr = NULL; 2537 2538 uint32_t sym_idx = 0; 2539 Symbol *sym = NULL; 2540 size_t num_syms = 0; 2541 std::string memory_symbol_name; 2542 uint32_t unmapped_local_symbols_found = 0; 2543 2544 std::vector<TrieEntryWithOffset> trie_entries; 2545 std::set<lldb::addr_t> resolver_addresses; 2546 2547 if (dyld_trie_data.GetByteSize() > 0) { 2548 std::vector<llvm::StringRef> nameSlices; 2549 ParseTrieEntries(dyld_trie_data, 0, is_arm, nameSlices, 2550 resolver_addresses, trie_entries); 2551 2552 ConstString text_segment_name("__TEXT"); 2553 SectionSP text_segment_sp = 2554 GetSectionList()->FindSectionByName(text_segment_name); 2555 if (text_segment_sp) { 2556 const lldb::addr_t text_segment_file_addr = 2557 text_segment_sp->GetFileAddress(); 2558 if (text_segment_file_addr != LLDB_INVALID_ADDRESS) { 2559 for (auto &e : trie_entries) 2560 e.entry.address += text_segment_file_addr; 2561 } 2562 } 2563 } 2564 2565 typedef std::set<ConstString> IndirectSymbols; 2566 IndirectSymbols indirect_symbol_names; 2567 2568 #if defined(__APPLE__) && \ 2569 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 2570 2571 // Some recent builds of the dyld_shared_cache (hereafter: DSC) have been 2572 // optimized by moving LOCAL symbols out of the memory mapped portion of 2573 // the DSC. The symbol information has all been retained, but it isn't 2574 // available in the normal nlist data. However, there *are* duplicate 2575 // entries of *some* 2576 // LOCAL symbols in the normal nlist data. To handle this situation 2577 // correctly, we must first attempt 2578 // to parse any DSC unmapped symbol information. If we find any, we set a 2579 // flag that tells the normal nlist parser to ignore all LOCAL symbols. 2580 2581 if (m_header.flags & 0x80000000u) { 2582 // Before we can start mapping the DSC, we need to make certain the 2583 // target process is actually using the cache we can find. 2584 2585 // Next we need to determine the correct path for the dyld shared cache. 2586 2587 ArchSpec header_arch; 2588 GetArchitecture(header_arch); 2589 char dsc_path[PATH_MAX]; 2590 char dsc_path_development[PATH_MAX]; 2591 2592 snprintf( 2593 dsc_path, sizeof(dsc_path), "%s%s%s", 2594 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2595 */ 2596 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2597 header_arch.GetArchitectureName()); 2598 2599 snprintf( 2600 dsc_path_development, sizeof(dsc_path), "%s%s%s%s", 2601 "/System/Library/Caches/com.apple.dyld/", /* IPHONE_DYLD_SHARED_CACHE_DIR 2602 */ 2603 "dyld_shared_cache_", /* DYLD_SHARED_CACHE_BASE_NAME */ 2604 header_arch.GetArchitectureName(), ".development"); 2605 2606 FileSpec dsc_nondevelopment_filespec(dsc_path, false); 2607 FileSpec dsc_development_filespec(dsc_path_development, false); 2608 FileSpec dsc_filespec; 2609 2610 UUID dsc_uuid; 2611 UUID process_shared_cache_uuid; 2612 addr_t process_shared_cache_base_addr; 2613 2614 if (process) { 2615 GetProcessSharedCacheUUID(process, process_shared_cache_base_addr, process_shared_cache_uuid); 2616 } 2617 2618 // First see if we can find an exact match for the inferior process 2619 // shared cache UUID in the development or non-development shared caches 2620 // on disk. 2621 if (process_shared_cache_uuid.IsValid()) { 2622 if (dsc_development_filespec.Exists()) { 2623 UUID dsc_development_uuid = GetSharedCacheUUID( 2624 dsc_development_filespec, byte_order, addr_byte_size); 2625 if (dsc_development_uuid.IsValid() && 2626 dsc_development_uuid == process_shared_cache_uuid) { 2627 dsc_filespec = dsc_development_filespec; 2628 dsc_uuid = dsc_development_uuid; 2629 } 2630 } 2631 if (!dsc_uuid.IsValid() && dsc_nondevelopment_filespec.Exists()) { 2632 UUID dsc_nondevelopment_uuid = GetSharedCacheUUID( 2633 dsc_nondevelopment_filespec, byte_order, addr_byte_size); 2634 if (dsc_nondevelopment_uuid.IsValid() && 2635 dsc_nondevelopment_uuid == process_shared_cache_uuid) { 2636 dsc_filespec = dsc_nondevelopment_filespec; 2637 dsc_uuid = dsc_nondevelopment_uuid; 2638 } 2639 } 2640 } 2641 2642 // Failing a UUID match, prefer the development dyld_shared cache if both 2643 // are present. 2644 if (!dsc_filespec.Exists()) { 2645 if (dsc_development_filespec.Exists()) { 2646 dsc_filespec = dsc_development_filespec; 2647 } else { 2648 dsc_filespec = dsc_nondevelopment_filespec; 2649 } 2650 } 2651 2652 /* The dyld_cache_header has a pointer to the 2653 dyld_cache_local_symbols_info structure (localSymbolsOffset). 2654 The dyld_cache_local_symbols_info structure gives us three things: 2655 1. The start and count of the nlist records in the dyld_shared_cache 2656 file 2657 2. The start and size of the strings for these nlist records 2658 3. The start and count of dyld_cache_local_symbols_entry entries 2659 2660 There is one dyld_cache_local_symbols_entry per dylib/framework in the 2661 dyld shared cache. 2662 The "dylibOffset" field is the Mach-O header of this dylib/framework in 2663 the dyld shared cache. 2664 The dyld_cache_local_symbols_entry also lists the start of this 2665 dylib/framework's nlist records 2666 and the count of how many nlist records there are for this 2667 dylib/framework. 2668 */ 2669 2670 // Process the dyld shared cache header to find the unmapped symbols 2671 2672 DataBufferSP dsc_data_sp = MapFileData( 2673 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_header_v1), 0); 2674 if (!dsc_uuid.IsValid()) { 2675 dsc_uuid = GetSharedCacheUUID(dsc_filespec, byte_order, addr_byte_size); 2676 } 2677 if (dsc_data_sp) { 2678 DataExtractor dsc_header_data(dsc_data_sp, byte_order, addr_byte_size); 2679 2680 bool uuid_match = true; 2681 if (dsc_uuid.IsValid() && process) { 2682 if (process_shared_cache_uuid.IsValid() && 2683 dsc_uuid != process_shared_cache_uuid) { 2684 // The on-disk dyld_shared_cache file is not the same as the one in 2685 // this process' memory, don't use it. 2686 uuid_match = false; 2687 ModuleSP module_sp(GetModule()); 2688 if (module_sp) 2689 module_sp->ReportWarning("process shared cache does not match " 2690 "on-disk dyld_shared_cache file, some " 2691 "symbol names will be missing."); 2692 } 2693 } 2694 2695 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, mappingOffset); 2696 2697 uint32_t mappingOffset = dsc_header_data.GetU32(&offset); 2698 2699 // If the mappingOffset points to a location inside the header, we've 2700 // opened an old dyld shared cache, and should not proceed further. 2701 if (uuid_match && 2702 mappingOffset >= sizeof(struct lldb_copy_dyld_cache_header_v1)) { 2703 2704 DataBufferSP dsc_mapping_info_data_sp = MapFileData( 2705 dsc_filespec, sizeof(struct lldb_copy_dyld_cache_mapping_info), 2706 mappingOffset); 2707 2708 DataExtractor dsc_mapping_info_data(dsc_mapping_info_data_sp, 2709 byte_order, addr_byte_size); 2710 offset = 0; 2711 2712 // The File addresses (from the in-memory Mach-O load commands) for 2713 // the shared libraries in the shared library cache need to be 2714 // adjusted by an offset to match up with the dylibOffset identifying 2715 // field in the dyld_cache_local_symbol_entry's. This offset is 2716 // recorded in mapping_offset_value. 2717 const uint64_t mapping_offset_value = 2718 dsc_mapping_info_data.GetU64(&offset); 2719 2720 offset = offsetof(struct lldb_copy_dyld_cache_header_v1, 2721 localSymbolsOffset); 2722 uint64_t localSymbolsOffset = dsc_header_data.GetU64(&offset); 2723 uint64_t localSymbolsSize = dsc_header_data.GetU64(&offset); 2724 2725 if (localSymbolsOffset && localSymbolsSize) { 2726 // Map the local symbols 2727 DataBufferSP dsc_local_symbols_data_sp = 2728 MapFileData(dsc_filespec, localSymbolsSize, localSymbolsOffset); 2729 2730 if (dsc_local_symbols_data_sp) { 2731 DataExtractor dsc_local_symbols_data(dsc_local_symbols_data_sp, 2732 byte_order, addr_byte_size); 2733 2734 offset = 0; 2735 2736 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 2737 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 2738 UndefinedNameToDescMap undefined_name_to_desc; 2739 SymbolIndexToName reexport_shlib_needs_fixup; 2740 2741 // Read the local_symbols_infos struct in one shot 2742 struct lldb_copy_dyld_cache_local_symbols_info local_symbols_info; 2743 dsc_local_symbols_data.GetU32(&offset, 2744 &local_symbols_info.nlistOffset, 6); 2745 2746 SectionSP text_section_sp( 2747 section_list->FindSectionByName(GetSegmentNameTEXT())); 2748 2749 uint32_t header_file_offset = 2750 (text_section_sp->GetFileAddress() - mapping_offset_value); 2751 2752 offset = local_symbols_info.entriesOffset; 2753 for (uint32_t entry_index = 0; 2754 entry_index < local_symbols_info.entriesCount; 2755 entry_index++) { 2756 struct lldb_copy_dyld_cache_local_symbols_entry 2757 local_symbols_entry; 2758 local_symbols_entry.dylibOffset = 2759 dsc_local_symbols_data.GetU32(&offset); 2760 local_symbols_entry.nlistStartIndex = 2761 dsc_local_symbols_data.GetU32(&offset); 2762 local_symbols_entry.nlistCount = 2763 dsc_local_symbols_data.GetU32(&offset); 2764 2765 if (header_file_offset == local_symbols_entry.dylibOffset) { 2766 unmapped_local_symbols_found = local_symbols_entry.nlistCount; 2767 2768 // The normal nlist code cannot correctly size the Symbols 2769 // array, we need to allocate it here. 2770 sym = symtab->Resize( 2771 symtab_load_command.nsyms + m_dysymtab.nindirectsyms + 2772 unmapped_local_symbols_found - m_dysymtab.nlocalsym); 2773 num_syms = symtab->GetNumSymbols(); 2774 2775 nlist_data_offset = 2776 local_symbols_info.nlistOffset + 2777 (nlist_byte_size * local_symbols_entry.nlistStartIndex); 2778 uint32_t string_table_offset = 2779 local_symbols_info.stringsOffset; 2780 2781 for (uint32_t nlist_index = 0; 2782 nlist_index < local_symbols_entry.nlistCount; 2783 nlist_index++) { 2784 ///////////////////////////// 2785 { 2786 struct nlist_64 nlist; 2787 if (!dsc_local_symbols_data.ValidOffsetForDataOfSize( 2788 nlist_data_offset, nlist_byte_size)) 2789 break; 2790 2791 nlist.n_strx = dsc_local_symbols_data.GetU32_unchecked( 2792 &nlist_data_offset); 2793 nlist.n_type = dsc_local_symbols_data.GetU8_unchecked( 2794 &nlist_data_offset); 2795 nlist.n_sect = dsc_local_symbols_data.GetU8_unchecked( 2796 &nlist_data_offset); 2797 nlist.n_desc = dsc_local_symbols_data.GetU16_unchecked( 2798 &nlist_data_offset); 2799 nlist.n_value = 2800 dsc_local_symbols_data.GetAddress_unchecked( 2801 &nlist_data_offset); 2802 2803 SymbolType type = eSymbolTypeInvalid; 2804 const char *symbol_name = dsc_local_symbols_data.PeekCStr( 2805 string_table_offset + nlist.n_strx); 2806 2807 if (symbol_name == NULL) { 2808 // No symbol should be NULL, even the symbols with no 2809 // string values should have an offset zero which 2810 // points to an empty C-string 2811 Host::SystemLog( 2812 Host::eSystemLogError, 2813 "error: DSC unmapped local symbol[%u] has invalid " 2814 "string table offset 0x%x in %s, ignoring symbol\n", 2815 entry_index, nlist.n_strx, 2816 module_sp->GetFileSpec().GetPath().c_str()); 2817 continue; 2818 } 2819 if (symbol_name[0] == '\0') 2820 symbol_name = NULL; 2821 2822 const char *symbol_name_non_abi_mangled = NULL; 2823 2824 SectionSP symbol_section; 2825 uint32_t symbol_byte_size = 0; 2826 bool add_nlist = true; 2827 bool is_debug = ((nlist.n_type & N_STAB) != 0); 2828 bool demangled_is_synthesized = false; 2829 bool is_gsym = false; 2830 bool set_value = true; 2831 2832 assert(sym_idx < num_syms); 2833 2834 sym[sym_idx].SetDebug(is_debug); 2835 2836 if (is_debug) { 2837 switch (nlist.n_type) { 2838 case N_GSYM: 2839 // global symbol: name,,NO_SECT,type,0 2840 // Sometimes the N_GSYM value contains the address. 2841 2842 // FIXME: In the .o files, we have a GSYM and a debug 2843 // symbol for all the ObjC data. They 2844 // have the same address, but we want to ensure that 2845 // we always find only the real symbol, 'cause we 2846 // don't currently correctly attribute the 2847 // GSYM one to the ObjCClass/Ivar/MetaClass 2848 // symbol type. This is a temporary hack to make 2849 // sure the ObjectiveC symbols get treated correctly. 2850 // To do this right, we should coalesce all the GSYM 2851 // & global symbols that have the same address. 2852 2853 is_gsym = true; 2854 sym[sym_idx].SetExternal(true); 2855 2856 if (symbol_name && symbol_name[0] == '_' && 2857 symbol_name[1] == 'O') { 2858 llvm::StringRef symbol_name_ref(symbol_name); 2859 if (symbol_name_ref.startswith( 2860 g_objc_v2_prefix_class)) { 2861 symbol_name_non_abi_mangled = symbol_name + 1; 2862 symbol_name = 2863 symbol_name + g_objc_v2_prefix_class.size(); 2864 type = eSymbolTypeObjCClass; 2865 demangled_is_synthesized = true; 2866 2867 } else if (symbol_name_ref.startswith( 2868 g_objc_v2_prefix_metaclass)) { 2869 symbol_name_non_abi_mangled = symbol_name + 1; 2870 symbol_name = symbol_name + 2871 g_objc_v2_prefix_metaclass.size(); 2872 type = eSymbolTypeObjCMetaClass; 2873 demangled_is_synthesized = true; 2874 } else if (symbol_name_ref.startswith( 2875 g_objc_v2_prefix_ivar)) { 2876 symbol_name_non_abi_mangled = symbol_name + 1; 2877 symbol_name = 2878 symbol_name + g_objc_v2_prefix_ivar.size(); 2879 type = eSymbolTypeObjCIVar; 2880 demangled_is_synthesized = true; 2881 } 2882 } else { 2883 if (nlist.n_value != 0) 2884 symbol_section = section_info.GetSection( 2885 nlist.n_sect, nlist.n_value); 2886 type = eSymbolTypeData; 2887 } 2888 break; 2889 2890 case N_FNAME: 2891 // procedure name (f77 kludge): name,,NO_SECT,0,0 2892 type = eSymbolTypeCompiler; 2893 break; 2894 2895 case N_FUN: 2896 // procedure: name,,n_sect,linenumber,address 2897 if (symbol_name) { 2898 type = eSymbolTypeCode; 2899 symbol_section = section_info.GetSection( 2900 nlist.n_sect, nlist.n_value); 2901 2902 N_FUN_addr_to_sym_idx.insert( 2903 std::make_pair(nlist.n_value, sym_idx)); 2904 // We use the current number of symbols in the 2905 // symbol table in lieu of using nlist_idx in case 2906 // we ever start trimming entries out 2907 N_FUN_indexes.push_back(sym_idx); 2908 } else { 2909 type = eSymbolTypeCompiler; 2910 2911 if (!N_FUN_indexes.empty()) { 2912 // Copy the size of the function into the 2913 // original 2914 // STAB entry so we don't have 2915 // to hunt for it later 2916 symtab->SymbolAtIndex(N_FUN_indexes.back()) 2917 ->SetByteSize(nlist.n_value); 2918 N_FUN_indexes.pop_back(); 2919 // We don't really need the end function STAB as 2920 // it contains the size which we already placed 2921 // with the original symbol, so don't add it if 2922 // we want a minimal symbol table 2923 add_nlist = false; 2924 } 2925 } 2926 break; 2927 2928 case N_STSYM: 2929 // static symbol: name,,n_sect,type,address 2930 N_STSYM_addr_to_sym_idx.insert( 2931 std::make_pair(nlist.n_value, sym_idx)); 2932 symbol_section = section_info.GetSection( 2933 nlist.n_sect, nlist.n_value); 2934 if (symbol_name && symbol_name[0]) { 2935 type = ObjectFile::GetSymbolTypeFromName( 2936 symbol_name + 1, eSymbolTypeData); 2937 } 2938 break; 2939 2940 case N_LCSYM: 2941 // .lcomm symbol: name,,n_sect,type,address 2942 symbol_section = section_info.GetSection( 2943 nlist.n_sect, nlist.n_value); 2944 type = eSymbolTypeCommonBlock; 2945 break; 2946 2947 case N_BNSYM: 2948 // We use the current number of symbols in the symbol 2949 // table in lieu of using nlist_idx in case we ever 2950 // start trimming entries out Skip these if we want 2951 // minimal symbol tables 2952 add_nlist = false; 2953 break; 2954 2955 case N_ENSYM: 2956 // Set the size of the N_BNSYM to the terminating 2957 // index of this N_ENSYM so that we can always skip 2958 // the entire symbol if we need to navigate more 2959 // quickly at the source level when parsing STABS 2960 // Skip these if we want minimal symbol tables 2961 add_nlist = false; 2962 break; 2963 2964 case N_OPT: 2965 // emitted with gcc2_compiled and in gcc source 2966 type = eSymbolTypeCompiler; 2967 break; 2968 2969 case N_RSYM: 2970 // register sym: name,,NO_SECT,type,register 2971 type = eSymbolTypeVariable; 2972 break; 2973 2974 case N_SLINE: 2975 // src line: 0,,n_sect,linenumber,address 2976 symbol_section = section_info.GetSection( 2977 nlist.n_sect, nlist.n_value); 2978 type = eSymbolTypeLineEntry; 2979 break; 2980 2981 case N_SSYM: 2982 // structure elt: name,,NO_SECT,type,struct_offset 2983 type = eSymbolTypeVariableType; 2984 break; 2985 2986 case N_SO: 2987 // source file name 2988 type = eSymbolTypeSourceFile; 2989 if (symbol_name == NULL) { 2990 add_nlist = false; 2991 if (N_SO_index != UINT32_MAX) { 2992 // Set the size of the N_SO to the terminating 2993 // index of this N_SO so that we can always skip 2994 // the entire N_SO if we need to navigate more 2995 // quickly at the source level when parsing STABS 2996 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 2997 symbol_ptr->SetByteSize(sym_idx); 2998 symbol_ptr->SetSizeIsSibling(true); 2999 } 3000 N_NSYM_indexes.clear(); 3001 N_INCL_indexes.clear(); 3002 N_BRAC_indexes.clear(); 3003 N_COMM_indexes.clear(); 3004 N_FUN_indexes.clear(); 3005 N_SO_index = UINT32_MAX; 3006 } else { 3007 // We use the current number of symbols in the 3008 // symbol table in lieu of using nlist_idx in case 3009 // we ever start trimming entries out 3010 const bool N_SO_has_full_path = 3011 symbol_name[0] == '/'; 3012 if (N_SO_has_full_path) { 3013 if ((N_SO_index == sym_idx - 1) && 3014 ((sym_idx - 1) < num_syms)) { 3015 // We have two consecutive N_SO entries where 3016 // the first contains a directory and the 3017 // second contains a full path. 3018 sym[sym_idx - 1].GetMangled().SetValue( 3019 ConstString(symbol_name), false); 3020 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3021 add_nlist = false; 3022 } else { 3023 // This is the first entry in a N_SO that 3024 // contains a directory or 3025 // a full path to the source file 3026 N_SO_index = sym_idx; 3027 } 3028 } else if ((N_SO_index == sym_idx - 1) && 3029 ((sym_idx - 1) < num_syms)) { 3030 // This is usually the second N_SO entry that 3031 // contains just the filename, so here we combine 3032 // it with the first one if we are minimizing the 3033 // symbol table 3034 const char *so_path = 3035 sym[sym_idx - 1] 3036 .GetMangled() 3037 .GetDemangledName( 3038 lldb::eLanguageTypeUnknown) 3039 .AsCString(); 3040 if (so_path && so_path[0]) { 3041 std::string full_so_path(so_path); 3042 const size_t double_slash_pos = 3043 full_so_path.find("//"); 3044 if (double_slash_pos != std::string::npos) { 3045 // The linker has been generating bad N_SO 3046 // entries with doubled up paths 3047 // in the format "%s%s" where the first 3048 // string in the DW_AT_comp_dir, and the 3049 // second is the directory for the source 3050 // file so you end up with a path that looks 3051 // like "/tmp/src//tmp/src/" 3052 FileSpec so_dir(so_path, false); 3053 if (!so_dir.Exists()) { 3054 so_dir.SetFile( 3055 &full_so_path[double_slash_pos + 1], 3056 false); 3057 if (so_dir.Exists()) { 3058 // Trim off the incorrect path 3059 full_so_path.erase(0, 3060 double_slash_pos + 1); 3061 } 3062 } 3063 } 3064 if (*full_so_path.rbegin() != '/') 3065 full_so_path += '/'; 3066 full_so_path += symbol_name; 3067 sym[sym_idx - 1].GetMangled().SetValue( 3068 ConstString(full_so_path.c_str()), false); 3069 add_nlist = false; 3070 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3071 } 3072 } else { 3073 // This could be a relative path to a N_SO 3074 N_SO_index = sym_idx; 3075 } 3076 } 3077 break; 3078 3079 case N_OSO: 3080 // object file name: name,,0,0,st_mtime 3081 type = eSymbolTypeObjectFile; 3082 break; 3083 3084 case N_LSYM: 3085 // local sym: name,,NO_SECT,type,offset 3086 type = eSymbolTypeLocal; 3087 break; 3088 3089 //---------------------------------------------------------------------- 3090 // INCL scopes 3091 //---------------------------------------------------------------------- 3092 case N_BINCL: 3093 // include file beginning: name,,NO_SECT,0,sum We use 3094 // the current number of symbols in the symbol table 3095 // in lieu of using nlist_idx in case we ever start 3096 // trimming entries out 3097 N_INCL_indexes.push_back(sym_idx); 3098 type = eSymbolTypeScopeBegin; 3099 break; 3100 3101 case N_EINCL: 3102 // include file end: name,,NO_SECT,0,0 3103 // Set the size of the N_BINCL to the terminating 3104 // index of this N_EINCL so that we can always skip 3105 // the entire symbol if we need to navigate more 3106 // quickly at the source level when parsing STABS 3107 if (!N_INCL_indexes.empty()) { 3108 symbol_ptr = 3109 symtab->SymbolAtIndex(N_INCL_indexes.back()); 3110 symbol_ptr->SetByteSize(sym_idx + 1); 3111 symbol_ptr->SetSizeIsSibling(true); 3112 N_INCL_indexes.pop_back(); 3113 } 3114 type = eSymbolTypeScopeEnd; 3115 break; 3116 3117 case N_SOL: 3118 // #included file name: name,,n_sect,0,address 3119 type = eSymbolTypeHeaderFile; 3120 3121 // We currently don't use the header files on darwin 3122 add_nlist = false; 3123 break; 3124 3125 case N_PARAMS: 3126 // compiler parameters: name,,NO_SECT,0,0 3127 type = eSymbolTypeCompiler; 3128 break; 3129 3130 case N_VERSION: 3131 // compiler version: name,,NO_SECT,0,0 3132 type = eSymbolTypeCompiler; 3133 break; 3134 3135 case N_OLEVEL: 3136 // compiler -O level: name,,NO_SECT,0,0 3137 type = eSymbolTypeCompiler; 3138 break; 3139 3140 case N_PSYM: 3141 // parameter: name,,NO_SECT,type,offset 3142 type = eSymbolTypeVariable; 3143 break; 3144 3145 case N_ENTRY: 3146 // alternate entry: name,,n_sect,linenumber,address 3147 symbol_section = section_info.GetSection( 3148 nlist.n_sect, nlist.n_value); 3149 type = eSymbolTypeLineEntry; 3150 break; 3151 3152 //---------------------------------------------------------------------- 3153 // Left and Right Braces 3154 //---------------------------------------------------------------------- 3155 case N_LBRAC: 3156 // left bracket: 0,,NO_SECT,nesting level,address We 3157 // use the current number of symbols in the symbol 3158 // table in lieu of using nlist_idx in case we ever 3159 // start trimming entries out 3160 symbol_section = section_info.GetSection( 3161 nlist.n_sect, nlist.n_value); 3162 N_BRAC_indexes.push_back(sym_idx); 3163 type = eSymbolTypeScopeBegin; 3164 break; 3165 3166 case N_RBRAC: 3167 // right bracket: 0,,NO_SECT,nesting level,address 3168 // Set the size of the N_LBRAC to the terminating 3169 // index of this N_RBRAC so that we can always skip 3170 // the entire symbol if we need to navigate more 3171 // quickly at the source level when parsing STABS 3172 symbol_section = section_info.GetSection( 3173 nlist.n_sect, nlist.n_value); 3174 if (!N_BRAC_indexes.empty()) { 3175 symbol_ptr = 3176 symtab->SymbolAtIndex(N_BRAC_indexes.back()); 3177 symbol_ptr->SetByteSize(sym_idx + 1); 3178 symbol_ptr->SetSizeIsSibling(true); 3179 N_BRAC_indexes.pop_back(); 3180 } 3181 type = eSymbolTypeScopeEnd; 3182 break; 3183 3184 case N_EXCL: 3185 // deleted include file: name,,NO_SECT,0,sum 3186 type = eSymbolTypeHeaderFile; 3187 break; 3188 3189 //---------------------------------------------------------------------- 3190 // COMM scopes 3191 //---------------------------------------------------------------------- 3192 case N_BCOMM: 3193 // begin common: name,,NO_SECT,0,0 3194 // We use the current number of symbols in the symbol 3195 // table in lieu of using nlist_idx in case we ever 3196 // start trimming entries out 3197 type = eSymbolTypeScopeBegin; 3198 N_COMM_indexes.push_back(sym_idx); 3199 break; 3200 3201 case N_ECOML: 3202 // end common (local name): 0,,n_sect,0,address 3203 symbol_section = section_info.GetSection( 3204 nlist.n_sect, nlist.n_value); 3205 // Fall through 3206 3207 case N_ECOMM: 3208 // end common: name,,n_sect,0,0 3209 // Set the size of the N_BCOMM to the terminating 3210 // index of this N_ECOMM/N_ECOML so that we can 3211 // always skip the entire symbol if we need to 3212 // navigate more quickly at the source level when 3213 // parsing STABS 3214 if (!N_COMM_indexes.empty()) { 3215 symbol_ptr = 3216 symtab->SymbolAtIndex(N_COMM_indexes.back()); 3217 symbol_ptr->SetByteSize(sym_idx + 1); 3218 symbol_ptr->SetSizeIsSibling(true); 3219 N_COMM_indexes.pop_back(); 3220 } 3221 type = eSymbolTypeScopeEnd; 3222 break; 3223 3224 case N_LENG: 3225 // second stab entry with length information 3226 type = eSymbolTypeAdditional; 3227 break; 3228 3229 default: 3230 break; 3231 } 3232 } else { 3233 // uint8_t n_pext = N_PEXT & nlist.n_type; 3234 uint8_t n_type = N_TYPE & nlist.n_type; 3235 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 3236 3237 switch (n_type) { 3238 case N_INDR: { 3239 const char *reexport_name_cstr = 3240 strtab_data.PeekCStr(nlist.n_value); 3241 if (reexport_name_cstr && reexport_name_cstr[0]) { 3242 type = eSymbolTypeReExported; 3243 ConstString reexport_name( 3244 reexport_name_cstr + 3245 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 3246 sym[sym_idx].SetReExportedSymbolName(reexport_name); 3247 set_value = false; 3248 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 3249 indirect_symbol_names.insert( 3250 ConstString(symbol_name + 3251 ((symbol_name[0] == '_') ? 1 : 0))); 3252 } else 3253 type = eSymbolTypeUndefined; 3254 } break; 3255 3256 case N_UNDF: 3257 if (symbol_name && symbol_name[0]) { 3258 ConstString undefined_name( 3259 symbol_name + 3260 ((symbol_name[0] == '_') ? 1 : 0)); 3261 undefined_name_to_desc[undefined_name] = 3262 nlist.n_desc; 3263 } 3264 // Fall through 3265 case N_PBUD: 3266 type = eSymbolTypeUndefined; 3267 break; 3268 3269 case N_ABS: 3270 type = eSymbolTypeAbsolute; 3271 break; 3272 3273 case N_SECT: { 3274 symbol_section = section_info.GetSection( 3275 nlist.n_sect, nlist.n_value); 3276 3277 if (symbol_section == NULL) { 3278 // TODO: warn about this? 3279 add_nlist = false; 3280 break; 3281 } 3282 3283 if (TEXT_eh_frame_sectID == nlist.n_sect) { 3284 type = eSymbolTypeException; 3285 } else { 3286 uint32_t section_type = 3287 symbol_section->Get() & SECTION_TYPE; 3288 3289 switch (section_type) { 3290 case S_CSTRING_LITERALS: 3291 type = eSymbolTypeData; 3292 break; // section with only literal C strings 3293 case S_4BYTE_LITERALS: 3294 type = eSymbolTypeData; 3295 break; // section with only 4 byte literals 3296 case S_8BYTE_LITERALS: 3297 type = eSymbolTypeData; 3298 break; // section with only 8 byte literals 3299 case S_LITERAL_POINTERS: 3300 type = eSymbolTypeTrampoline; 3301 break; // section with only pointers to literals 3302 case S_NON_LAZY_SYMBOL_POINTERS: 3303 type = eSymbolTypeTrampoline; 3304 break; // section with only non-lazy symbol 3305 // pointers 3306 case S_LAZY_SYMBOL_POINTERS: 3307 type = eSymbolTypeTrampoline; 3308 break; // section with only lazy symbol pointers 3309 case S_SYMBOL_STUBS: 3310 type = eSymbolTypeTrampoline; 3311 break; // section with only symbol stubs, byte 3312 // size of stub in the reserved2 field 3313 case S_MOD_INIT_FUNC_POINTERS: 3314 type = eSymbolTypeCode; 3315 break; // section with only function pointers for 3316 // initialization 3317 case S_MOD_TERM_FUNC_POINTERS: 3318 type = eSymbolTypeCode; 3319 break; // section with only function pointers for 3320 // termination 3321 case S_INTERPOSING: 3322 type = eSymbolTypeTrampoline; 3323 break; // section with only pairs of function 3324 // pointers for interposing 3325 case S_16BYTE_LITERALS: 3326 type = eSymbolTypeData; 3327 break; // section with only 16 byte literals 3328 case S_DTRACE_DOF: 3329 type = eSymbolTypeInstrumentation; 3330 break; 3331 case S_LAZY_DYLIB_SYMBOL_POINTERS: 3332 type = eSymbolTypeTrampoline; 3333 break; 3334 default: 3335 switch (symbol_section->GetType()) { 3336 case lldb::eSectionTypeCode: 3337 type = eSymbolTypeCode; 3338 break; 3339 case eSectionTypeData: 3340 case eSectionTypeDataCString: // Inlined C string 3341 // data 3342 case eSectionTypeDataCStringPointers: // Pointers 3343 // to C 3344 // string 3345 // data 3346 case eSectionTypeDataSymbolAddress: // Address of 3347 // a symbol in 3348 // the symbol 3349 // table 3350 case eSectionTypeData4: 3351 case eSectionTypeData8: 3352 case eSectionTypeData16: 3353 type = eSymbolTypeData; 3354 break; 3355 default: 3356 break; 3357 } 3358 break; 3359 } 3360 3361 if (type == eSymbolTypeInvalid) { 3362 const char *symbol_sect_name = 3363 symbol_section->GetName().AsCString(); 3364 if (symbol_section->IsDescendant( 3365 text_section_sp.get())) { 3366 if (symbol_section->IsClear( 3367 S_ATTR_PURE_INSTRUCTIONS | 3368 S_ATTR_SELF_MODIFYING_CODE | 3369 S_ATTR_SOME_INSTRUCTIONS)) 3370 type = eSymbolTypeData; 3371 else 3372 type = eSymbolTypeCode; 3373 } else if (symbol_section->IsDescendant( 3374 data_section_sp.get()) || 3375 symbol_section->IsDescendant( 3376 data_dirty_section_sp.get()) || 3377 symbol_section->IsDescendant( 3378 data_const_section_sp.get())) { 3379 if (symbol_sect_name && 3380 ::strstr(symbol_sect_name, "__objc") == 3381 symbol_sect_name) { 3382 type = eSymbolTypeRuntime; 3383 3384 if (symbol_name) { 3385 llvm::StringRef symbol_name_ref( 3386 symbol_name); 3387 if (symbol_name_ref.startswith("_OBJC_")) { 3388 static const llvm::StringRef 3389 g_objc_v2_prefix_class( 3390 "_OBJC_CLASS_$_"); 3391 static const llvm::StringRef 3392 g_objc_v2_prefix_metaclass( 3393 "_OBJC_METACLASS_$_"); 3394 static const llvm::StringRef 3395 g_objc_v2_prefix_ivar( 3396 "_OBJC_IVAR_$_"); 3397 if (symbol_name_ref.startswith( 3398 g_objc_v2_prefix_class)) { 3399 symbol_name_non_abi_mangled = 3400 symbol_name + 1; 3401 symbol_name = 3402 symbol_name + 3403 g_objc_v2_prefix_class.size(); 3404 type = eSymbolTypeObjCClass; 3405 demangled_is_synthesized = true; 3406 } else if ( 3407 symbol_name_ref.startswith( 3408 g_objc_v2_prefix_metaclass)) { 3409 symbol_name_non_abi_mangled = 3410 symbol_name + 1; 3411 symbol_name = 3412 symbol_name + 3413 g_objc_v2_prefix_metaclass.size(); 3414 type = eSymbolTypeObjCMetaClass; 3415 demangled_is_synthesized = true; 3416 } else if (symbol_name_ref.startswith( 3417 g_objc_v2_prefix_ivar)) { 3418 symbol_name_non_abi_mangled = 3419 symbol_name + 1; 3420 symbol_name = 3421 symbol_name + 3422 g_objc_v2_prefix_ivar.size(); 3423 type = eSymbolTypeObjCIVar; 3424 demangled_is_synthesized = true; 3425 } 3426 } 3427 } 3428 } else if (symbol_sect_name && 3429 ::strstr(symbol_sect_name, 3430 "__gcc_except_tab") == 3431 symbol_sect_name) { 3432 type = eSymbolTypeException; 3433 } else { 3434 type = eSymbolTypeData; 3435 } 3436 } else if (symbol_sect_name && 3437 ::strstr(symbol_sect_name, 3438 "__IMPORT") == 3439 symbol_sect_name) { 3440 type = eSymbolTypeTrampoline; 3441 } else if (symbol_section->IsDescendant( 3442 objc_section_sp.get())) { 3443 type = eSymbolTypeRuntime; 3444 if (symbol_name && symbol_name[0] == '.') { 3445 llvm::StringRef symbol_name_ref(symbol_name); 3446 static const llvm::StringRef 3447 g_objc_v1_prefix_class( 3448 ".objc_class_name_"); 3449 if (symbol_name_ref.startswith( 3450 g_objc_v1_prefix_class)) { 3451 symbol_name_non_abi_mangled = symbol_name; 3452 symbol_name = symbol_name + 3453 g_objc_v1_prefix_class.size(); 3454 type = eSymbolTypeObjCClass; 3455 demangled_is_synthesized = true; 3456 } 3457 } 3458 } 3459 } 3460 } 3461 } break; 3462 } 3463 } 3464 3465 if (add_nlist) { 3466 uint64_t symbol_value = nlist.n_value; 3467 if (symbol_name_non_abi_mangled) { 3468 sym[sym_idx].GetMangled().SetMangledName( 3469 ConstString(symbol_name_non_abi_mangled)); 3470 sym[sym_idx].GetMangled().SetDemangledName( 3471 ConstString(symbol_name)); 3472 } else { 3473 bool symbol_name_is_mangled = false; 3474 3475 if (symbol_name && symbol_name[0] == '_') { 3476 symbol_name_is_mangled = symbol_name[1] == '_'; 3477 symbol_name++; // Skip the leading underscore 3478 } 3479 3480 if (symbol_name) { 3481 ConstString const_symbol_name(symbol_name); 3482 sym[sym_idx].GetMangled().SetValue( 3483 const_symbol_name, symbol_name_is_mangled); 3484 if (is_gsym && is_debug) { 3485 const char *gsym_name = 3486 sym[sym_idx] 3487 .GetMangled() 3488 .GetName(lldb::eLanguageTypeUnknown, 3489 Mangled::ePreferMangled) 3490 .GetCString(); 3491 if (gsym_name) 3492 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 3493 } 3494 } 3495 } 3496 if (symbol_section) { 3497 const addr_t section_file_addr = 3498 symbol_section->GetFileAddress(); 3499 if (symbol_byte_size == 0 && 3500 function_starts_count > 0) { 3501 addr_t symbol_lookup_file_addr = nlist.n_value; 3502 // Do an exact address match for non-ARM addresses, 3503 // else get the closest since the symbol might be a 3504 // thumb symbol which has an address with bit zero 3505 // set 3506 FunctionStarts::Entry *func_start_entry = 3507 function_starts.FindEntry( 3508 symbol_lookup_file_addr, !is_arm); 3509 if (is_arm && func_start_entry) { 3510 // Verify that the function start address is the 3511 // symbol address (ARM) or the symbol address + 1 3512 // (thumb) 3513 if (func_start_entry->addr != 3514 symbol_lookup_file_addr && 3515 func_start_entry->addr != 3516 (symbol_lookup_file_addr + 1)) { 3517 // Not the right entry, NULL it out... 3518 func_start_entry = NULL; 3519 } 3520 } 3521 if (func_start_entry) { 3522 func_start_entry->data = true; 3523 3524 addr_t symbol_file_addr = func_start_entry->addr; 3525 uint32_t symbol_flags = 0; 3526 if (is_arm) { 3527 if (symbol_file_addr & 1) 3528 symbol_flags = 3529 MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 3530 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 3531 } 3532 3533 const FunctionStarts::Entry 3534 *next_func_start_entry = 3535 function_starts.FindNextEntry( 3536 func_start_entry); 3537 const addr_t section_end_file_addr = 3538 section_file_addr + 3539 symbol_section->GetByteSize(); 3540 if (next_func_start_entry) { 3541 addr_t next_symbol_file_addr = 3542 next_func_start_entry->addr; 3543 // Be sure the clear the Thumb address bit when 3544 // we calculate the size from the current and 3545 // next address 3546 if (is_arm) 3547 next_symbol_file_addr &= 3548 THUMB_ADDRESS_BIT_MASK; 3549 symbol_byte_size = std::min<lldb::addr_t>( 3550 next_symbol_file_addr - symbol_file_addr, 3551 section_end_file_addr - symbol_file_addr); 3552 } else { 3553 symbol_byte_size = 3554 section_end_file_addr - symbol_file_addr; 3555 } 3556 } 3557 } 3558 symbol_value -= section_file_addr; 3559 } 3560 3561 if (is_debug == false) { 3562 if (type == eSymbolTypeCode) { 3563 // See if we can find a N_FUN entry for any code 3564 // symbols. If we do find a match, and the name 3565 // matches, then we can merge the two into just the 3566 // function symbol to avoid duplicate entries in 3567 // the symbol table 3568 std::pair<ValueToSymbolIndexMap::const_iterator, 3569 ValueToSymbolIndexMap::const_iterator> 3570 range; 3571 range = N_FUN_addr_to_sym_idx.equal_range( 3572 nlist.n_value); 3573 if (range.first != range.second) { 3574 bool found_it = false; 3575 for (ValueToSymbolIndexMap::const_iterator pos = 3576 range.first; 3577 pos != range.second; ++pos) { 3578 if (sym[sym_idx].GetMangled().GetName( 3579 lldb::eLanguageTypeUnknown, 3580 Mangled::ePreferMangled) == 3581 sym[pos->second].GetMangled().GetName( 3582 lldb::eLanguageTypeUnknown, 3583 Mangled::ePreferMangled)) { 3584 m_nlist_idx_to_sym_idx[nlist_idx] = 3585 pos->second; 3586 // We just need the flags from the linker 3587 // symbol, so put these flags 3588 // into the N_FUN flags to avoid duplicate 3589 // symbols in the symbol table 3590 sym[pos->second].SetExternal( 3591 sym[sym_idx].IsExternal()); 3592 sym[pos->second].SetFlags(nlist.n_type << 16 | 3593 nlist.n_desc); 3594 if (resolver_addresses.find(nlist.n_value) != 3595 resolver_addresses.end()) 3596 sym[pos->second].SetType( 3597 eSymbolTypeResolver); 3598 sym[sym_idx].Clear(); 3599 found_it = true; 3600 break; 3601 } 3602 } 3603 if (found_it) 3604 continue; 3605 } else { 3606 if (resolver_addresses.find(nlist.n_value) != 3607 resolver_addresses.end()) 3608 type = eSymbolTypeResolver; 3609 } 3610 } else if (type == eSymbolTypeData || 3611 type == eSymbolTypeObjCClass || 3612 type == eSymbolTypeObjCMetaClass || 3613 type == eSymbolTypeObjCIVar) { 3614 // See if we can find a N_STSYM entry for any data 3615 // symbols. If we do find a match, and the name 3616 // matches, then we can merge the two into just the 3617 // Static symbol to avoid duplicate entries in the 3618 // symbol table 3619 std::pair<ValueToSymbolIndexMap::const_iterator, 3620 ValueToSymbolIndexMap::const_iterator> 3621 range; 3622 range = N_STSYM_addr_to_sym_idx.equal_range( 3623 nlist.n_value); 3624 if (range.first != range.second) { 3625 bool found_it = false; 3626 for (ValueToSymbolIndexMap::const_iterator pos = 3627 range.first; 3628 pos != range.second; ++pos) { 3629 if (sym[sym_idx].GetMangled().GetName( 3630 lldb::eLanguageTypeUnknown, 3631 Mangled::ePreferMangled) == 3632 sym[pos->second].GetMangled().GetName( 3633 lldb::eLanguageTypeUnknown, 3634 Mangled::ePreferMangled)) { 3635 m_nlist_idx_to_sym_idx[nlist_idx] = 3636 pos->second; 3637 // We just need the flags from the linker 3638 // symbol, so put these flags 3639 // into the N_STSYM flags to avoid duplicate 3640 // symbols in the symbol table 3641 sym[pos->second].SetExternal( 3642 sym[sym_idx].IsExternal()); 3643 sym[pos->second].SetFlags(nlist.n_type << 16 | 3644 nlist.n_desc); 3645 sym[sym_idx].Clear(); 3646 found_it = true; 3647 break; 3648 } 3649 } 3650 if (found_it) 3651 continue; 3652 } else { 3653 const char *gsym_name = 3654 sym[sym_idx] 3655 .GetMangled() 3656 .GetName(lldb::eLanguageTypeUnknown, 3657 Mangled::ePreferMangled) 3658 .GetCString(); 3659 if (gsym_name) { 3660 // Combine N_GSYM stab entries with the non 3661 // stab symbol 3662 ConstNameToSymbolIndexMap::const_iterator pos = 3663 N_GSYM_name_to_sym_idx.find(gsym_name); 3664 if (pos != N_GSYM_name_to_sym_idx.end()) { 3665 const uint32_t GSYM_sym_idx = pos->second; 3666 m_nlist_idx_to_sym_idx[nlist_idx] = 3667 GSYM_sym_idx; 3668 // Copy the address, because often the N_GSYM 3669 // address has an invalid address of zero 3670 // when the global is a common symbol 3671 sym[GSYM_sym_idx].GetAddressRef().SetSection( 3672 symbol_section); 3673 sym[GSYM_sym_idx].GetAddressRef().SetOffset( 3674 symbol_value); 3675 // We just need the flags from the linker 3676 // symbol, so put these flags 3677 // into the N_GSYM flags to avoid duplicate 3678 // symbols in the symbol table 3679 sym[GSYM_sym_idx].SetFlags( 3680 nlist.n_type << 16 | nlist.n_desc); 3681 sym[sym_idx].Clear(); 3682 continue; 3683 } 3684 } 3685 } 3686 } 3687 } 3688 3689 sym[sym_idx].SetID(nlist_idx); 3690 sym[sym_idx].SetType(type); 3691 if (set_value) { 3692 sym[sym_idx].GetAddressRef().SetSection( 3693 symbol_section); 3694 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 3695 } 3696 sym[sym_idx].SetFlags(nlist.n_type << 16 | 3697 nlist.n_desc); 3698 3699 if (symbol_byte_size > 0) 3700 sym[sym_idx].SetByteSize(symbol_byte_size); 3701 3702 if (demangled_is_synthesized) 3703 sym[sym_idx].SetDemangledNameIsSynthesized(true); 3704 ++sym_idx; 3705 } else { 3706 sym[sym_idx].Clear(); 3707 } 3708 } 3709 ///////////////////////////// 3710 } 3711 break; // No more entries to consider 3712 } 3713 } 3714 3715 for (const auto &pos : reexport_shlib_needs_fixup) { 3716 const auto undef_pos = undefined_name_to_desc.find(pos.second); 3717 if (undef_pos != undefined_name_to_desc.end()) { 3718 const uint8_t dylib_ordinal = 3719 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 3720 if (dylib_ordinal > 0 && 3721 dylib_ordinal < dylib_files.GetSize()) 3722 sym[pos.first].SetReExportedSymbolSharedLibrary( 3723 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 3724 } 3725 } 3726 } 3727 } 3728 } 3729 } 3730 } 3731 3732 // Must reset this in case it was mutated above! 3733 nlist_data_offset = 0; 3734 #endif 3735 3736 if (nlist_data.GetByteSize() > 0) { 3737 3738 // If the sym array was not created while parsing the DSC unmapped 3739 // symbols, create it now. 3740 if (sym == NULL) { 3741 sym = symtab->Resize(symtab_load_command.nsyms + 3742 m_dysymtab.nindirectsyms); 3743 num_syms = symtab->GetNumSymbols(); 3744 } 3745 3746 if (unmapped_local_symbols_found) { 3747 assert(m_dysymtab.ilocalsym == 0); 3748 nlist_data_offset += (m_dysymtab.nlocalsym * nlist_byte_size); 3749 nlist_idx = m_dysymtab.nlocalsym; 3750 } else { 3751 nlist_idx = 0; 3752 } 3753 3754 typedef std::map<ConstString, uint16_t> UndefinedNameToDescMap; 3755 typedef std::map<uint32_t, ConstString> SymbolIndexToName; 3756 UndefinedNameToDescMap undefined_name_to_desc; 3757 SymbolIndexToName reexport_shlib_needs_fixup; 3758 for (; nlist_idx < symtab_load_command.nsyms; ++nlist_idx) { 3759 struct nlist_64 nlist; 3760 if (!nlist_data.ValidOffsetForDataOfSize(nlist_data_offset, 3761 nlist_byte_size)) 3762 break; 3763 3764 nlist.n_strx = nlist_data.GetU32_unchecked(&nlist_data_offset); 3765 nlist.n_type = nlist_data.GetU8_unchecked(&nlist_data_offset); 3766 nlist.n_sect = nlist_data.GetU8_unchecked(&nlist_data_offset); 3767 nlist.n_desc = nlist_data.GetU16_unchecked(&nlist_data_offset); 3768 nlist.n_value = nlist_data.GetAddress_unchecked(&nlist_data_offset); 3769 3770 SymbolType type = eSymbolTypeInvalid; 3771 const char *symbol_name = NULL; 3772 3773 if (have_strtab_data) { 3774 symbol_name = strtab_data.PeekCStr(nlist.n_strx); 3775 3776 if (symbol_name == NULL) { 3777 // No symbol should be NULL, even the symbols with no string values 3778 // should have an offset zero which points to an empty C-string 3779 Host::SystemLog(Host::eSystemLogError, 3780 "error: symbol[%u] has invalid string table offset " 3781 "0x%x in %s, ignoring symbol\n", 3782 nlist_idx, nlist.n_strx, 3783 module_sp->GetFileSpec().GetPath().c_str()); 3784 continue; 3785 } 3786 if (symbol_name[0] == '\0') 3787 symbol_name = NULL; 3788 } else { 3789 const addr_t str_addr = strtab_addr + nlist.n_strx; 3790 Status str_error; 3791 if (process->ReadCStringFromMemory(str_addr, memory_symbol_name, 3792 str_error)) 3793 symbol_name = memory_symbol_name.c_str(); 3794 } 3795 const char *symbol_name_non_abi_mangled = NULL; 3796 3797 SectionSP symbol_section; 3798 lldb::addr_t symbol_byte_size = 0; 3799 bool add_nlist = true; 3800 bool is_gsym = false; 3801 bool is_debug = ((nlist.n_type & N_STAB) != 0); 3802 bool demangled_is_synthesized = false; 3803 bool set_value = true; 3804 assert(sym_idx < num_syms); 3805 3806 sym[sym_idx].SetDebug(is_debug); 3807 3808 if (is_debug) { 3809 switch (nlist.n_type) { 3810 case N_GSYM: 3811 // global symbol: name,,NO_SECT,type,0 3812 // Sometimes the N_GSYM value contains the address. 3813 3814 // FIXME: In the .o files, we have a GSYM and a debug symbol for all 3815 // the ObjC data. They 3816 // have the same address, but we want to ensure that we always find 3817 // only the real symbol, 'cause we don't currently correctly 3818 // attribute the GSYM one to the ObjCClass/Ivar/MetaClass symbol 3819 // type. This is a temporary hack to make sure the ObjectiveC 3820 // symbols get treated correctly. To do this right, we should 3821 // coalesce all the GSYM & global symbols that have the same 3822 // address. 3823 is_gsym = true; 3824 sym[sym_idx].SetExternal(true); 3825 3826 if (symbol_name && symbol_name[0] == '_' && symbol_name[1] == 'O') { 3827 llvm::StringRef symbol_name_ref(symbol_name); 3828 if (symbol_name_ref.startswith(g_objc_v2_prefix_class)) { 3829 symbol_name_non_abi_mangled = symbol_name + 1; 3830 symbol_name = symbol_name + g_objc_v2_prefix_class.size(); 3831 type = eSymbolTypeObjCClass; 3832 demangled_is_synthesized = true; 3833 3834 } else if (symbol_name_ref.startswith( 3835 g_objc_v2_prefix_metaclass)) { 3836 symbol_name_non_abi_mangled = symbol_name + 1; 3837 symbol_name = symbol_name + g_objc_v2_prefix_metaclass.size(); 3838 type = eSymbolTypeObjCMetaClass; 3839 demangled_is_synthesized = true; 3840 } else if (symbol_name_ref.startswith(g_objc_v2_prefix_ivar)) { 3841 symbol_name_non_abi_mangled = symbol_name + 1; 3842 symbol_name = symbol_name + g_objc_v2_prefix_ivar.size(); 3843 type = eSymbolTypeObjCIVar; 3844 demangled_is_synthesized = true; 3845 } 3846 } else { 3847 if (nlist.n_value != 0) 3848 symbol_section = 3849 section_info.GetSection(nlist.n_sect, nlist.n_value); 3850 type = eSymbolTypeData; 3851 } 3852 break; 3853 3854 case N_FNAME: 3855 // procedure name (f77 kludge): name,,NO_SECT,0,0 3856 type = eSymbolTypeCompiler; 3857 break; 3858 3859 case N_FUN: 3860 // procedure: name,,n_sect,linenumber,address 3861 if (symbol_name) { 3862 type = eSymbolTypeCode; 3863 symbol_section = 3864 section_info.GetSection(nlist.n_sect, nlist.n_value); 3865 3866 N_FUN_addr_to_sym_idx.insert( 3867 std::make_pair(nlist.n_value, sym_idx)); 3868 // We use the current number of symbols in the symbol table in 3869 // lieu of using nlist_idx in case we ever start trimming entries 3870 // out 3871 N_FUN_indexes.push_back(sym_idx); 3872 } else { 3873 type = eSymbolTypeCompiler; 3874 3875 if (!N_FUN_indexes.empty()) { 3876 // Copy the size of the function into the original STAB entry 3877 // so we don't have to hunt for it later 3878 symtab->SymbolAtIndex(N_FUN_indexes.back()) 3879 ->SetByteSize(nlist.n_value); 3880 N_FUN_indexes.pop_back(); 3881 // We don't really need the end function STAB as it contains 3882 // the size which we already placed with the original symbol, 3883 // so don't add it if we want a minimal symbol table 3884 add_nlist = false; 3885 } 3886 } 3887 break; 3888 3889 case N_STSYM: 3890 // static symbol: name,,n_sect,type,address 3891 N_STSYM_addr_to_sym_idx.insert( 3892 std::make_pair(nlist.n_value, sym_idx)); 3893 symbol_section = 3894 section_info.GetSection(nlist.n_sect, nlist.n_value); 3895 if (symbol_name && symbol_name[0]) { 3896 type = ObjectFile::GetSymbolTypeFromName(symbol_name + 1, 3897 eSymbolTypeData); 3898 } 3899 break; 3900 3901 case N_LCSYM: 3902 // .lcomm symbol: name,,n_sect,type,address 3903 symbol_section = 3904 section_info.GetSection(nlist.n_sect, nlist.n_value); 3905 type = eSymbolTypeCommonBlock; 3906 break; 3907 3908 case N_BNSYM: 3909 // We use the current number of symbols in the symbol table in lieu 3910 // of using nlist_idx in case we ever start trimming entries out 3911 // Skip these if we want minimal symbol tables 3912 add_nlist = false; 3913 break; 3914 3915 case N_ENSYM: 3916 // Set the size of the N_BNSYM to the terminating index of this 3917 // N_ENSYM so that we can always skip the entire symbol if we need 3918 // to navigate more quickly at the source level when parsing STABS 3919 // Skip these if we want minimal symbol tables 3920 add_nlist = false; 3921 break; 3922 3923 case N_OPT: 3924 // emitted with gcc2_compiled and in gcc source 3925 type = eSymbolTypeCompiler; 3926 break; 3927 3928 case N_RSYM: 3929 // register sym: name,,NO_SECT,type,register 3930 type = eSymbolTypeVariable; 3931 break; 3932 3933 case N_SLINE: 3934 // src line: 0,,n_sect,linenumber,address 3935 symbol_section = 3936 section_info.GetSection(nlist.n_sect, nlist.n_value); 3937 type = eSymbolTypeLineEntry; 3938 break; 3939 3940 case N_SSYM: 3941 // structure elt: name,,NO_SECT,type,struct_offset 3942 type = eSymbolTypeVariableType; 3943 break; 3944 3945 case N_SO: 3946 // source file name 3947 type = eSymbolTypeSourceFile; 3948 if (symbol_name == NULL) { 3949 add_nlist = false; 3950 if (N_SO_index != UINT32_MAX) { 3951 // Set the size of the N_SO to the terminating index of this 3952 // N_SO so that we can always skip the entire N_SO if we need 3953 // to navigate more quickly at the source level when parsing 3954 // STABS 3955 symbol_ptr = symtab->SymbolAtIndex(N_SO_index); 3956 symbol_ptr->SetByteSize(sym_idx); 3957 symbol_ptr->SetSizeIsSibling(true); 3958 } 3959 N_NSYM_indexes.clear(); 3960 N_INCL_indexes.clear(); 3961 N_BRAC_indexes.clear(); 3962 N_COMM_indexes.clear(); 3963 N_FUN_indexes.clear(); 3964 N_SO_index = UINT32_MAX; 3965 } else { 3966 // We use the current number of symbols in the symbol table in 3967 // lieu of using nlist_idx in case we ever start trimming entries 3968 // out 3969 const bool N_SO_has_full_path = symbol_name[0] == '/'; 3970 if (N_SO_has_full_path) { 3971 if ((N_SO_index == sym_idx - 1) && ((sym_idx - 1) < num_syms)) { 3972 // We have two consecutive N_SO entries where the first 3973 // contains a directory and the second contains a full path. 3974 sym[sym_idx - 1].GetMangled().SetValue( 3975 ConstString(symbol_name), false); 3976 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 3977 add_nlist = false; 3978 } else { 3979 // This is the first entry in a N_SO that contains a 3980 // directory or a full path to the source file 3981 N_SO_index = sym_idx; 3982 } 3983 } else if ((N_SO_index == sym_idx - 1) && 3984 ((sym_idx - 1) < num_syms)) { 3985 // This is usually the second N_SO entry that contains just the 3986 // filename, so here we combine it with the first one if we are 3987 // minimizing the symbol table 3988 const char *so_path = 3989 sym[sym_idx - 1] 3990 .GetMangled() 3991 .GetDemangledName(lldb::eLanguageTypeUnknown) 3992 .AsCString(); 3993 if (so_path && so_path[0]) { 3994 std::string full_so_path(so_path); 3995 const size_t double_slash_pos = full_so_path.find("//"); 3996 if (double_slash_pos != std::string::npos) { 3997 // The linker has been generating bad N_SO entries with 3998 // doubled up paths in the format "%s%s" where the first 3999 // string in the DW_AT_comp_dir, and the second is the 4000 // directory for the source file so you end up with a path 4001 // that looks like "/tmp/src//tmp/src/" 4002 FileSpec so_dir(so_path, false); 4003 if (!so_dir.Exists()) { 4004 so_dir.SetFile(&full_so_path[double_slash_pos + 1], 4005 false); 4006 if (so_dir.Exists()) { 4007 // Trim off the incorrect path 4008 full_so_path.erase(0, double_slash_pos + 1); 4009 } 4010 } 4011 } 4012 if (*full_so_path.rbegin() != '/') 4013 full_so_path += '/'; 4014 full_so_path += symbol_name; 4015 sym[sym_idx - 1].GetMangled().SetValue( 4016 ConstString(full_so_path.c_str()), false); 4017 add_nlist = false; 4018 m_nlist_idx_to_sym_idx[nlist_idx] = sym_idx - 1; 4019 } 4020 } else { 4021 // This could be a relative path to a N_SO 4022 N_SO_index = sym_idx; 4023 } 4024 } 4025 break; 4026 4027 case N_OSO: 4028 // object file name: name,,0,0,st_mtime 4029 type = eSymbolTypeObjectFile; 4030 break; 4031 4032 case N_LSYM: 4033 // local sym: name,,NO_SECT,type,offset 4034 type = eSymbolTypeLocal; 4035 break; 4036 4037 //---------------------------------------------------------------------- 4038 // INCL scopes 4039 //---------------------------------------------------------------------- 4040 case N_BINCL: 4041 // include file beginning: name,,NO_SECT,0,sum We use the current 4042 // number of symbols in the symbol table in lieu of using nlist_idx 4043 // in case we ever start trimming entries out 4044 N_INCL_indexes.push_back(sym_idx); 4045 type = eSymbolTypeScopeBegin; 4046 break; 4047 4048 case N_EINCL: 4049 // include file end: name,,NO_SECT,0,0 4050 // Set the size of the N_BINCL to the terminating index of this 4051 // N_EINCL so that we can always skip the entire symbol if we need 4052 // to navigate more quickly at the source level when parsing STABS 4053 if (!N_INCL_indexes.empty()) { 4054 symbol_ptr = symtab->SymbolAtIndex(N_INCL_indexes.back()); 4055 symbol_ptr->SetByteSize(sym_idx + 1); 4056 symbol_ptr->SetSizeIsSibling(true); 4057 N_INCL_indexes.pop_back(); 4058 } 4059 type = eSymbolTypeScopeEnd; 4060 break; 4061 4062 case N_SOL: 4063 // #included file name: name,,n_sect,0,address 4064 type = eSymbolTypeHeaderFile; 4065 4066 // We currently don't use the header files on darwin 4067 add_nlist = false; 4068 break; 4069 4070 case N_PARAMS: 4071 // compiler parameters: name,,NO_SECT,0,0 4072 type = eSymbolTypeCompiler; 4073 break; 4074 4075 case N_VERSION: 4076 // compiler version: name,,NO_SECT,0,0 4077 type = eSymbolTypeCompiler; 4078 break; 4079 4080 case N_OLEVEL: 4081 // compiler -O level: name,,NO_SECT,0,0 4082 type = eSymbolTypeCompiler; 4083 break; 4084 4085 case N_PSYM: 4086 // parameter: name,,NO_SECT,type,offset 4087 type = eSymbolTypeVariable; 4088 break; 4089 4090 case N_ENTRY: 4091 // alternate entry: name,,n_sect,linenumber,address 4092 symbol_section = 4093 section_info.GetSection(nlist.n_sect, nlist.n_value); 4094 type = eSymbolTypeLineEntry; 4095 break; 4096 4097 //---------------------------------------------------------------------- 4098 // Left and Right Braces 4099 //---------------------------------------------------------------------- 4100 case N_LBRAC: 4101 // left bracket: 0,,NO_SECT,nesting level,address We use the 4102 // current number of symbols in the symbol table in lieu of using 4103 // nlist_idx in case we ever start trimming entries out 4104 symbol_section = 4105 section_info.GetSection(nlist.n_sect, nlist.n_value); 4106 N_BRAC_indexes.push_back(sym_idx); 4107 type = eSymbolTypeScopeBegin; 4108 break; 4109 4110 case N_RBRAC: 4111 // right bracket: 0,,NO_SECT,nesting level,address Set the size of 4112 // the N_LBRAC to the terminating index of this N_RBRAC so that we 4113 // can always skip the entire symbol if we need to navigate more 4114 // quickly at the source level when parsing STABS 4115 symbol_section = 4116 section_info.GetSection(nlist.n_sect, nlist.n_value); 4117 if (!N_BRAC_indexes.empty()) { 4118 symbol_ptr = symtab->SymbolAtIndex(N_BRAC_indexes.back()); 4119 symbol_ptr->SetByteSize(sym_idx + 1); 4120 symbol_ptr->SetSizeIsSibling(true); 4121 N_BRAC_indexes.pop_back(); 4122 } 4123 type = eSymbolTypeScopeEnd; 4124 break; 4125 4126 case N_EXCL: 4127 // deleted include file: name,,NO_SECT,0,sum 4128 type = eSymbolTypeHeaderFile; 4129 break; 4130 4131 //---------------------------------------------------------------------- 4132 // COMM scopes 4133 //---------------------------------------------------------------------- 4134 case N_BCOMM: 4135 // begin common: name,,NO_SECT,0,0 4136 // We use the current number of symbols in the symbol table in lieu 4137 // of using nlist_idx in case we ever start trimming entries out 4138 type = eSymbolTypeScopeBegin; 4139 N_COMM_indexes.push_back(sym_idx); 4140 break; 4141 4142 case N_ECOML: 4143 // end common (local name): 0,,n_sect,0,address 4144 symbol_section = 4145 section_info.GetSection(nlist.n_sect, nlist.n_value); 4146 LLVM_FALLTHROUGH; 4147 4148 case N_ECOMM: 4149 // end common: name,,n_sect,0,0 4150 // Set the size of the N_BCOMM to the terminating index of this 4151 // N_ECOMM/N_ECOML so that we can always skip the entire symbol if 4152 // we need to navigate more quickly at the source level when 4153 // parsing STABS 4154 if (!N_COMM_indexes.empty()) { 4155 symbol_ptr = symtab->SymbolAtIndex(N_COMM_indexes.back()); 4156 symbol_ptr->SetByteSize(sym_idx + 1); 4157 symbol_ptr->SetSizeIsSibling(true); 4158 N_COMM_indexes.pop_back(); 4159 } 4160 type = eSymbolTypeScopeEnd; 4161 break; 4162 4163 case N_LENG: 4164 // second stab entry with length information 4165 type = eSymbolTypeAdditional; 4166 break; 4167 4168 default: 4169 break; 4170 } 4171 } else { 4172 // uint8_t n_pext = N_PEXT & nlist.n_type; 4173 uint8_t n_type = N_TYPE & nlist.n_type; 4174 sym[sym_idx].SetExternal((N_EXT & nlist.n_type) != 0); 4175 4176 switch (n_type) { 4177 case N_INDR: { 4178 const char *reexport_name_cstr = 4179 strtab_data.PeekCStr(nlist.n_value); 4180 if (reexport_name_cstr && reexport_name_cstr[0]) { 4181 type = eSymbolTypeReExported; 4182 ConstString reexport_name( 4183 reexport_name_cstr + 4184 ((reexport_name_cstr[0] == '_') ? 1 : 0)); 4185 sym[sym_idx].SetReExportedSymbolName(reexport_name); 4186 set_value = false; 4187 reexport_shlib_needs_fixup[sym_idx] = reexport_name; 4188 indirect_symbol_names.insert( 4189 ConstString(symbol_name + ((symbol_name[0] == '_') ? 1 : 0))); 4190 } else 4191 type = eSymbolTypeUndefined; 4192 } break; 4193 4194 case N_UNDF: 4195 if (symbol_name && symbol_name[0]) { 4196 ConstString undefined_name(symbol_name + 4197 ((symbol_name[0] == '_') ? 1 : 0)); 4198 undefined_name_to_desc[undefined_name] = nlist.n_desc; 4199 } 4200 LLVM_FALLTHROUGH; 4201 4202 case N_PBUD: 4203 type = eSymbolTypeUndefined; 4204 break; 4205 4206 case N_ABS: 4207 type = eSymbolTypeAbsolute; 4208 break; 4209 4210 case N_SECT: { 4211 symbol_section = 4212 section_info.GetSection(nlist.n_sect, nlist.n_value); 4213 4214 if (!symbol_section) { 4215 // TODO: warn about this? 4216 add_nlist = false; 4217 break; 4218 } 4219 4220 if (TEXT_eh_frame_sectID == nlist.n_sect) { 4221 type = eSymbolTypeException; 4222 } else { 4223 uint32_t section_type = symbol_section->Get() & SECTION_TYPE; 4224 4225 switch (section_type) { 4226 case S_CSTRING_LITERALS: 4227 type = eSymbolTypeData; 4228 break; // section with only literal C strings 4229 case S_4BYTE_LITERALS: 4230 type = eSymbolTypeData; 4231 break; // section with only 4 byte literals 4232 case S_8BYTE_LITERALS: 4233 type = eSymbolTypeData; 4234 break; // section with only 8 byte literals 4235 case S_LITERAL_POINTERS: 4236 type = eSymbolTypeTrampoline; 4237 break; // section with only pointers to literals 4238 case S_NON_LAZY_SYMBOL_POINTERS: 4239 type = eSymbolTypeTrampoline; 4240 break; // section with only non-lazy symbol pointers 4241 case S_LAZY_SYMBOL_POINTERS: 4242 type = eSymbolTypeTrampoline; 4243 break; // section with only lazy symbol pointers 4244 case S_SYMBOL_STUBS: 4245 type = eSymbolTypeTrampoline; 4246 break; // section with only symbol stubs, byte size of stub in 4247 // the reserved2 field 4248 case S_MOD_INIT_FUNC_POINTERS: 4249 type = eSymbolTypeCode; 4250 break; // section with only function pointers for initialization 4251 case S_MOD_TERM_FUNC_POINTERS: 4252 type = eSymbolTypeCode; 4253 break; // section with only function pointers for termination 4254 case S_INTERPOSING: 4255 type = eSymbolTypeTrampoline; 4256 break; // section with only pairs of function pointers for 4257 // interposing 4258 case S_16BYTE_LITERALS: 4259 type = eSymbolTypeData; 4260 break; // section with only 16 byte literals 4261 case S_DTRACE_DOF: 4262 type = eSymbolTypeInstrumentation; 4263 break; 4264 case S_LAZY_DYLIB_SYMBOL_POINTERS: 4265 type = eSymbolTypeTrampoline; 4266 break; 4267 default: 4268 switch (symbol_section->GetType()) { 4269 case lldb::eSectionTypeCode: 4270 type = eSymbolTypeCode; 4271 break; 4272 case eSectionTypeData: 4273 case eSectionTypeDataCString: // Inlined C string data 4274 case eSectionTypeDataCStringPointers: // Pointers to C string 4275 // data 4276 case eSectionTypeDataSymbolAddress: // Address of a symbol in 4277 // the symbol table 4278 case eSectionTypeData4: 4279 case eSectionTypeData8: 4280 case eSectionTypeData16: 4281 type = eSymbolTypeData; 4282 break; 4283 default: 4284 break; 4285 } 4286 break; 4287 } 4288 4289 if (type == eSymbolTypeInvalid) { 4290 const char *symbol_sect_name = 4291 symbol_section->GetName().AsCString(); 4292 if (symbol_section->IsDescendant(text_section_sp.get())) { 4293 if (symbol_section->IsClear(S_ATTR_PURE_INSTRUCTIONS | 4294 S_ATTR_SELF_MODIFYING_CODE | 4295 S_ATTR_SOME_INSTRUCTIONS)) 4296 type = eSymbolTypeData; 4297 else 4298 type = eSymbolTypeCode; 4299 } else if (symbol_section->IsDescendant( 4300 data_section_sp.get()) || 4301 symbol_section->IsDescendant( 4302 data_dirty_section_sp.get()) || 4303 symbol_section->IsDescendant( 4304 data_const_section_sp.get())) { 4305 if (symbol_sect_name && 4306 ::strstr(symbol_sect_name, "__objc") == 4307 symbol_sect_name) { 4308 type = eSymbolTypeRuntime; 4309 4310 if (symbol_name) { 4311 llvm::StringRef symbol_name_ref(symbol_name); 4312 if (symbol_name_ref.startswith("_OBJC_")) { 4313 static const llvm::StringRef g_objc_v2_prefix_class( 4314 "_OBJC_CLASS_$_"); 4315 static const llvm::StringRef g_objc_v2_prefix_metaclass( 4316 "_OBJC_METACLASS_$_"); 4317 static const llvm::StringRef g_objc_v2_prefix_ivar( 4318 "_OBJC_IVAR_$_"); 4319 if (symbol_name_ref.startswith( 4320 g_objc_v2_prefix_class)) { 4321 symbol_name_non_abi_mangled = symbol_name + 1; 4322 symbol_name = 4323 symbol_name + g_objc_v2_prefix_class.size(); 4324 type = eSymbolTypeObjCClass; 4325 demangled_is_synthesized = true; 4326 } else if (symbol_name_ref.startswith( 4327 g_objc_v2_prefix_metaclass)) { 4328 symbol_name_non_abi_mangled = symbol_name + 1; 4329 symbol_name = 4330 symbol_name + g_objc_v2_prefix_metaclass.size(); 4331 type = eSymbolTypeObjCMetaClass; 4332 demangled_is_synthesized = true; 4333 } else if (symbol_name_ref.startswith( 4334 g_objc_v2_prefix_ivar)) { 4335 symbol_name_non_abi_mangled = symbol_name + 1; 4336 symbol_name = 4337 symbol_name + g_objc_v2_prefix_ivar.size(); 4338 type = eSymbolTypeObjCIVar; 4339 demangled_is_synthesized = true; 4340 } 4341 } 4342 } 4343 } else if (symbol_sect_name && 4344 ::strstr(symbol_sect_name, "__gcc_except_tab") == 4345 symbol_sect_name) { 4346 type = eSymbolTypeException; 4347 } else { 4348 type = eSymbolTypeData; 4349 } 4350 } else if (symbol_sect_name && 4351 ::strstr(symbol_sect_name, "__IMPORT") == 4352 symbol_sect_name) { 4353 type = eSymbolTypeTrampoline; 4354 } else if (symbol_section->IsDescendant( 4355 objc_section_sp.get())) { 4356 type = eSymbolTypeRuntime; 4357 if (symbol_name && symbol_name[0] == '.') { 4358 llvm::StringRef symbol_name_ref(symbol_name); 4359 static const llvm::StringRef g_objc_v1_prefix_class( 4360 ".objc_class_name_"); 4361 if (symbol_name_ref.startswith(g_objc_v1_prefix_class)) { 4362 symbol_name_non_abi_mangled = symbol_name; 4363 symbol_name = symbol_name + g_objc_v1_prefix_class.size(); 4364 type = eSymbolTypeObjCClass; 4365 demangled_is_synthesized = true; 4366 } 4367 } 4368 } 4369 } 4370 } 4371 } break; 4372 } 4373 } 4374 4375 if (add_nlist) { 4376 uint64_t symbol_value = nlist.n_value; 4377 4378 if (symbol_name_non_abi_mangled) { 4379 sym[sym_idx].GetMangled().SetMangledName( 4380 ConstString(symbol_name_non_abi_mangled)); 4381 sym[sym_idx].GetMangled().SetDemangledName( 4382 ConstString(symbol_name)); 4383 } else { 4384 bool symbol_name_is_mangled = false; 4385 4386 if (symbol_name && symbol_name[0] == '_') { 4387 symbol_name_is_mangled = symbol_name[1] == '_'; 4388 symbol_name++; // Skip the leading underscore 4389 } 4390 4391 if (symbol_name) { 4392 ConstString const_symbol_name(symbol_name); 4393 sym[sym_idx].GetMangled().SetValue(const_symbol_name, 4394 symbol_name_is_mangled); 4395 } 4396 } 4397 4398 if (is_gsym) { 4399 const char *gsym_name = sym[sym_idx] 4400 .GetMangled() 4401 .GetName(lldb::eLanguageTypeUnknown, 4402 Mangled::ePreferMangled) 4403 .GetCString(); 4404 if (gsym_name) 4405 N_GSYM_name_to_sym_idx[gsym_name] = sym_idx; 4406 } 4407 4408 if (symbol_section) { 4409 const addr_t section_file_addr = symbol_section->GetFileAddress(); 4410 if (symbol_byte_size == 0 && function_starts_count > 0) { 4411 addr_t symbol_lookup_file_addr = nlist.n_value; 4412 // Do an exact address match for non-ARM addresses, else get the 4413 // closest since the symbol might be a thumb symbol which has an 4414 // address with bit zero set 4415 FunctionStarts::Entry *func_start_entry = 4416 function_starts.FindEntry(symbol_lookup_file_addr, !is_arm); 4417 if (is_arm && func_start_entry) { 4418 // Verify that the function start address is the symbol address 4419 // (ARM) or the symbol address + 1 (thumb) 4420 if (func_start_entry->addr != symbol_lookup_file_addr && 4421 func_start_entry->addr != (symbol_lookup_file_addr + 1)) { 4422 // Not the right entry, NULL it out... 4423 func_start_entry = NULL; 4424 } 4425 } 4426 if (func_start_entry) { 4427 func_start_entry->data = true; 4428 4429 addr_t symbol_file_addr = func_start_entry->addr; 4430 if (is_arm) 4431 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4432 4433 const FunctionStarts::Entry *next_func_start_entry = 4434 function_starts.FindNextEntry(func_start_entry); 4435 const addr_t section_end_file_addr = 4436 section_file_addr + symbol_section->GetByteSize(); 4437 if (next_func_start_entry) { 4438 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4439 // Be sure the clear the Thumb address bit when we calculate 4440 // the size from the current and next address 4441 if (is_arm) 4442 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4443 symbol_byte_size = std::min<lldb::addr_t>( 4444 next_symbol_file_addr - symbol_file_addr, 4445 section_end_file_addr - symbol_file_addr); 4446 } else { 4447 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4448 } 4449 } 4450 } 4451 symbol_value -= section_file_addr; 4452 } 4453 4454 if (is_debug == false) { 4455 if (type == eSymbolTypeCode) { 4456 // See if we can find a N_FUN entry for any code symbols. If we 4457 // do find a match, and the name matches, then we can merge the 4458 // two into just the function symbol to avoid duplicate entries 4459 // in the symbol table 4460 std::pair<ValueToSymbolIndexMap::const_iterator, 4461 ValueToSymbolIndexMap::const_iterator> 4462 range; 4463 range = N_FUN_addr_to_sym_idx.equal_range(nlist.n_value); 4464 if (range.first != range.second) { 4465 bool found_it = false; 4466 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4467 pos != range.second; ++pos) { 4468 if (sym[sym_idx].GetMangled().GetName( 4469 lldb::eLanguageTypeUnknown, 4470 Mangled::ePreferMangled) == 4471 sym[pos->second].GetMangled().GetName( 4472 lldb::eLanguageTypeUnknown, 4473 Mangled::ePreferMangled)) { 4474 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4475 // We just need the flags from the linker symbol, so put 4476 // these flags into the N_FUN flags to avoid duplicate 4477 // symbols in the symbol table 4478 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4479 sym[pos->second].SetFlags(nlist.n_type << 16 | 4480 nlist.n_desc); 4481 if (resolver_addresses.find(nlist.n_value) != 4482 resolver_addresses.end()) 4483 sym[pos->second].SetType(eSymbolTypeResolver); 4484 sym[sym_idx].Clear(); 4485 found_it = true; 4486 break; 4487 } 4488 } 4489 if (found_it) 4490 continue; 4491 } else { 4492 if (resolver_addresses.find(nlist.n_value) != 4493 resolver_addresses.end()) 4494 type = eSymbolTypeResolver; 4495 } 4496 } else if (type == eSymbolTypeData || 4497 type == eSymbolTypeObjCClass || 4498 type == eSymbolTypeObjCMetaClass || 4499 type == eSymbolTypeObjCIVar) { 4500 // See if we can find a N_STSYM entry for any data symbols. If we 4501 // do find a match, and the name matches, then we can merge the 4502 // two into just the Static symbol to avoid duplicate entries in 4503 // the symbol table 4504 std::pair<ValueToSymbolIndexMap::const_iterator, 4505 ValueToSymbolIndexMap::const_iterator> 4506 range; 4507 range = N_STSYM_addr_to_sym_idx.equal_range(nlist.n_value); 4508 if (range.first != range.second) { 4509 bool found_it = false; 4510 for (ValueToSymbolIndexMap::const_iterator pos = range.first; 4511 pos != range.second; ++pos) { 4512 if (sym[sym_idx].GetMangled().GetName( 4513 lldb::eLanguageTypeUnknown, 4514 Mangled::ePreferMangled) == 4515 sym[pos->second].GetMangled().GetName( 4516 lldb::eLanguageTypeUnknown, 4517 Mangled::ePreferMangled)) { 4518 m_nlist_idx_to_sym_idx[nlist_idx] = pos->second; 4519 // We just need the flags from the linker symbol, so put 4520 // these flags into the N_STSYM flags to avoid duplicate 4521 // symbols in the symbol table 4522 sym[pos->second].SetExternal(sym[sym_idx].IsExternal()); 4523 sym[pos->second].SetFlags(nlist.n_type << 16 | 4524 nlist.n_desc); 4525 sym[sym_idx].Clear(); 4526 found_it = true; 4527 break; 4528 } 4529 } 4530 if (found_it) 4531 continue; 4532 } else { 4533 // Combine N_GSYM stab entries with the non stab symbol 4534 const char *gsym_name = sym[sym_idx] 4535 .GetMangled() 4536 .GetName(lldb::eLanguageTypeUnknown, 4537 Mangled::ePreferMangled) 4538 .GetCString(); 4539 if (gsym_name) { 4540 ConstNameToSymbolIndexMap::const_iterator pos = 4541 N_GSYM_name_to_sym_idx.find(gsym_name); 4542 if (pos != N_GSYM_name_to_sym_idx.end()) { 4543 const uint32_t GSYM_sym_idx = pos->second; 4544 m_nlist_idx_to_sym_idx[nlist_idx] = GSYM_sym_idx; 4545 // Copy the address, because often the N_GSYM address has 4546 // an invalid address of zero when the global is a common 4547 // symbol 4548 sym[GSYM_sym_idx].GetAddressRef().SetSection( 4549 symbol_section); 4550 sym[GSYM_sym_idx].GetAddressRef().SetOffset(symbol_value); 4551 // We just need the flags from the linker symbol, so put 4552 // these flags into the N_GSYM flags to avoid duplicate 4553 // symbols in the symbol table 4554 sym[GSYM_sym_idx].SetFlags(nlist.n_type << 16 | 4555 nlist.n_desc); 4556 sym[sym_idx].Clear(); 4557 continue; 4558 } 4559 } 4560 } 4561 } 4562 } 4563 4564 sym[sym_idx].SetID(nlist_idx); 4565 sym[sym_idx].SetType(type); 4566 if (set_value) { 4567 sym[sym_idx].GetAddressRef().SetSection(symbol_section); 4568 sym[sym_idx].GetAddressRef().SetOffset(symbol_value); 4569 } 4570 sym[sym_idx].SetFlags(nlist.n_type << 16 | nlist.n_desc); 4571 4572 if (symbol_byte_size > 0) 4573 sym[sym_idx].SetByteSize(symbol_byte_size); 4574 4575 if (demangled_is_synthesized) 4576 sym[sym_idx].SetDemangledNameIsSynthesized(true); 4577 4578 ++sym_idx; 4579 } else { 4580 sym[sym_idx].Clear(); 4581 } 4582 } 4583 4584 for (const auto &pos : reexport_shlib_needs_fixup) { 4585 const auto undef_pos = undefined_name_to_desc.find(pos.second); 4586 if (undef_pos != undefined_name_to_desc.end()) { 4587 const uint8_t dylib_ordinal = 4588 llvm::MachO::GET_LIBRARY_ORDINAL(undef_pos->second); 4589 if (dylib_ordinal > 0 && dylib_ordinal < dylib_files.GetSize()) 4590 sym[pos.first].SetReExportedSymbolSharedLibrary( 4591 dylib_files.GetFileSpecAtIndex(dylib_ordinal - 1)); 4592 } 4593 } 4594 } 4595 4596 uint32_t synthetic_sym_id = symtab_load_command.nsyms; 4597 4598 if (function_starts_count > 0) { 4599 uint32_t num_synthetic_function_symbols = 0; 4600 for (i = 0; i < function_starts_count; ++i) { 4601 if (function_starts.GetEntryRef(i).data == false) 4602 ++num_synthetic_function_symbols; 4603 } 4604 4605 if (num_synthetic_function_symbols > 0) { 4606 if (num_syms < sym_idx + num_synthetic_function_symbols) { 4607 num_syms = sym_idx + num_synthetic_function_symbols; 4608 sym = symtab->Resize(num_syms); 4609 } 4610 for (i = 0; i < function_starts_count; ++i) { 4611 const FunctionStarts::Entry *func_start_entry = 4612 function_starts.GetEntryAtIndex(i); 4613 if (func_start_entry->data == false) { 4614 addr_t symbol_file_addr = func_start_entry->addr; 4615 uint32_t symbol_flags = 0; 4616 if (is_arm) { 4617 if (symbol_file_addr & 1) 4618 symbol_flags = MACHO_NLIST_ARM_SYMBOL_IS_THUMB; 4619 symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4620 } 4621 Address symbol_addr; 4622 if (module_sp->ResolveFileAddress(symbol_file_addr, symbol_addr)) { 4623 SectionSP symbol_section(symbol_addr.GetSection()); 4624 uint32_t symbol_byte_size = 0; 4625 if (symbol_section) { 4626 const addr_t section_file_addr = 4627 symbol_section->GetFileAddress(); 4628 const FunctionStarts::Entry *next_func_start_entry = 4629 function_starts.FindNextEntry(func_start_entry); 4630 const addr_t section_end_file_addr = 4631 section_file_addr + symbol_section->GetByteSize(); 4632 if (next_func_start_entry) { 4633 addr_t next_symbol_file_addr = next_func_start_entry->addr; 4634 if (is_arm) 4635 next_symbol_file_addr &= THUMB_ADDRESS_BIT_MASK; 4636 symbol_byte_size = std::min<lldb::addr_t>( 4637 next_symbol_file_addr - symbol_file_addr, 4638 section_end_file_addr - symbol_file_addr); 4639 } else { 4640 symbol_byte_size = section_end_file_addr - symbol_file_addr; 4641 } 4642 sym[sym_idx].SetID(synthetic_sym_id++); 4643 sym[sym_idx].GetMangled().SetDemangledName( 4644 GetNextSyntheticSymbolName()); 4645 sym[sym_idx].SetType(eSymbolTypeCode); 4646 sym[sym_idx].SetIsSynthetic(true); 4647 sym[sym_idx].GetAddressRef() = symbol_addr; 4648 if (symbol_flags) 4649 sym[sym_idx].SetFlags(symbol_flags); 4650 if (symbol_byte_size) 4651 sym[sym_idx].SetByteSize(symbol_byte_size); 4652 ++sym_idx; 4653 } 4654 } 4655 } 4656 } 4657 } 4658 } 4659 4660 // Trim our symbols down to just what we ended up with after removing any 4661 // symbols. 4662 if (sym_idx < num_syms) { 4663 num_syms = sym_idx; 4664 sym = symtab->Resize(num_syms); 4665 } 4666 4667 // Now synthesize indirect symbols 4668 if (m_dysymtab.nindirectsyms != 0) { 4669 if (indirect_symbol_index_data.GetByteSize()) { 4670 NListIndexToSymbolIndexMap::const_iterator end_index_pos = 4671 m_nlist_idx_to_sym_idx.end(); 4672 4673 for (uint32_t sect_idx = 1; sect_idx < m_mach_sections.size(); 4674 ++sect_idx) { 4675 if ((m_mach_sections[sect_idx].flags & SECTION_TYPE) == 4676 S_SYMBOL_STUBS) { 4677 uint32_t symbol_stub_byte_size = 4678 m_mach_sections[sect_idx].reserved2; 4679 if (symbol_stub_byte_size == 0) 4680 continue; 4681 4682 const uint32_t num_symbol_stubs = 4683 m_mach_sections[sect_idx].size / symbol_stub_byte_size; 4684 4685 if (num_symbol_stubs == 0) 4686 continue; 4687 4688 const uint32_t symbol_stub_index_offset = 4689 m_mach_sections[sect_idx].reserved1; 4690 for (uint32_t stub_idx = 0; stub_idx < num_symbol_stubs; 4691 ++stub_idx) { 4692 const uint32_t symbol_stub_index = 4693 symbol_stub_index_offset + stub_idx; 4694 const lldb::addr_t symbol_stub_addr = 4695 m_mach_sections[sect_idx].addr + 4696 (stub_idx * symbol_stub_byte_size); 4697 lldb::offset_t symbol_stub_offset = symbol_stub_index * 4; 4698 if (indirect_symbol_index_data.ValidOffsetForDataOfSize( 4699 symbol_stub_offset, 4)) { 4700 const uint32_t stub_sym_id = 4701 indirect_symbol_index_data.GetU32(&symbol_stub_offset); 4702 if (stub_sym_id & (INDIRECT_SYMBOL_ABS | INDIRECT_SYMBOL_LOCAL)) 4703 continue; 4704 4705 NListIndexToSymbolIndexMap::const_iterator index_pos = 4706 m_nlist_idx_to_sym_idx.find(stub_sym_id); 4707 Symbol *stub_symbol = NULL; 4708 if (index_pos != end_index_pos) { 4709 // We have a remapping from the original nlist index to a 4710 // current symbol index, so just look this up by index 4711 stub_symbol = symtab->SymbolAtIndex(index_pos->second); 4712 } else { 4713 // We need to lookup a symbol using the original nlist symbol 4714 // index since this index is coming from the S_SYMBOL_STUBS 4715 stub_symbol = symtab->FindSymbolByID(stub_sym_id); 4716 } 4717 4718 if (stub_symbol) { 4719 Address so_addr(symbol_stub_addr, section_list); 4720 4721 if (stub_symbol->GetType() == eSymbolTypeUndefined) { 4722 // Change the external symbol into a trampoline that makes 4723 // sense These symbols were N_UNDF N_EXT, and are useless 4724 // to us, so we can re-use them so we don't have to make up 4725 // a synthetic symbol for no good reason. 4726 if (resolver_addresses.find(symbol_stub_addr) == 4727 resolver_addresses.end()) 4728 stub_symbol->SetType(eSymbolTypeTrampoline); 4729 else 4730 stub_symbol->SetType(eSymbolTypeResolver); 4731 stub_symbol->SetExternal(false); 4732 stub_symbol->GetAddressRef() = so_addr; 4733 stub_symbol->SetByteSize(symbol_stub_byte_size); 4734 } else { 4735 // Make a synthetic symbol to describe the trampoline stub 4736 Mangled stub_symbol_mangled_name(stub_symbol->GetMangled()); 4737 if (sym_idx >= num_syms) { 4738 sym = symtab->Resize(++num_syms); 4739 stub_symbol = NULL; // this pointer no longer valid 4740 } 4741 sym[sym_idx].SetID(synthetic_sym_id++); 4742 sym[sym_idx].GetMangled() = stub_symbol_mangled_name; 4743 if (resolver_addresses.find(symbol_stub_addr) == 4744 resolver_addresses.end()) 4745 sym[sym_idx].SetType(eSymbolTypeTrampoline); 4746 else 4747 sym[sym_idx].SetType(eSymbolTypeResolver); 4748 sym[sym_idx].SetIsSynthetic(true); 4749 sym[sym_idx].GetAddressRef() = so_addr; 4750 sym[sym_idx].SetByteSize(symbol_stub_byte_size); 4751 ++sym_idx; 4752 } 4753 } else { 4754 if (log) 4755 log->Warning("symbol stub referencing symbol table symbol " 4756 "%u that isn't in our minimal symbol table, " 4757 "fix this!!!", 4758 stub_sym_id); 4759 } 4760 } 4761 } 4762 } 4763 } 4764 } 4765 } 4766 4767 if (!trie_entries.empty()) { 4768 for (const auto &e : trie_entries) { 4769 if (e.entry.import_name) { 4770 // Only add indirect symbols from the Trie entries if we didn't have 4771 // a N_INDR nlist entry for this already 4772 if (indirect_symbol_names.find(e.entry.name) == 4773 indirect_symbol_names.end()) { 4774 // Make a synthetic symbol to describe re-exported symbol. 4775 if (sym_idx >= num_syms) 4776 sym = symtab->Resize(++num_syms); 4777 sym[sym_idx].SetID(synthetic_sym_id++); 4778 sym[sym_idx].GetMangled() = Mangled(e.entry.name); 4779 sym[sym_idx].SetType(eSymbolTypeReExported); 4780 sym[sym_idx].SetIsSynthetic(true); 4781 sym[sym_idx].SetReExportedSymbolName(e.entry.import_name); 4782 if (e.entry.other > 0 && e.entry.other <= dylib_files.GetSize()) { 4783 sym[sym_idx].SetReExportedSymbolSharedLibrary( 4784 dylib_files.GetFileSpecAtIndex(e.entry.other - 1)); 4785 } 4786 ++sym_idx; 4787 } 4788 } 4789 } 4790 } 4791 4792 // StreamFile s(stdout, false); 4793 // s.Printf ("Symbol table before CalculateSymbolSizes():\n"); 4794 // symtab->Dump(&s, NULL, eSortOrderNone); 4795 // Set symbol byte sizes correctly since mach-o nlist entries don't have 4796 // sizes 4797 symtab->CalculateSymbolSizes(); 4798 4799 // s.Printf ("Symbol table after CalculateSymbolSizes():\n"); 4800 // symtab->Dump(&s, NULL, eSortOrderNone); 4801 4802 return symtab->GetNumSymbols(); 4803 } 4804 return 0; 4805 } 4806 4807 void ObjectFileMachO::Dump(Stream *s) { 4808 ModuleSP module_sp(GetModule()); 4809 if (module_sp) { 4810 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4811 s->Printf("%p: ", static_cast<void *>(this)); 4812 s->Indent(); 4813 if (m_header.magic == MH_MAGIC_64 || m_header.magic == MH_CIGAM_64) 4814 s->PutCString("ObjectFileMachO64"); 4815 else 4816 s->PutCString("ObjectFileMachO32"); 4817 4818 ArchSpec header_arch; 4819 GetArchitecture(header_arch); 4820 4821 *s << ", file = '" << m_file 4822 << "', triple = " << header_arch.GetTriple().getTriple() << "\n"; 4823 4824 SectionList *sections = GetSectionList(); 4825 if (sections) 4826 sections->Dump(s, NULL, true, UINT32_MAX); 4827 4828 if (m_symtab_ap.get()) 4829 m_symtab_ap->Dump(s, NULL, eSortOrderNone); 4830 } 4831 } 4832 4833 bool ObjectFileMachO::GetUUID(const llvm::MachO::mach_header &header, 4834 const lldb_private::DataExtractor &data, 4835 lldb::offset_t lc_offset, 4836 lldb_private::UUID &uuid) { 4837 uint32_t i; 4838 struct uuid_command load_cmd; 4839 4840 lldb::offset_t offset = lc_offset; 4841 for (i = 0; i < header.ncmds; ++i) { 4842 const lldb::offset_t cmd_offset = offset; 4843 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4844 break; 4845 4846 if (load_cmd.cmd == LC_UUID) { 4847 const uint8_t *uuid_bytes = data.PeekData(offset, 16); 4848 4849 if (uuid_bytes) { 4850 // OpenCL on Mac OS X uses the same UUID for each of its object files. 4851 // We pretend these object files have no UUID to prevent crashing. 4852 4853 const uint8_t opencl_uuid[] = {0x8c, 0x8e, 0xb3, 0x9b, 0x3b, 0xa8, 4854 0x4b, 0x16, 0xb6, 0xa4, 0x27, 0x63, 4855 0xbb, 0x14, 0xf0, 0x0d}; 4856 4857 if (!memcmp(uuid_bytes, opencl_uuid, 16)) 4858 return false; 4859 4860 uuid.SetBytes(uuid_bytes); 4861 return true; 4862 } 4863 return false; 4864 } 4865 offset = cmd_offset + load_cmd.cmdsize; 4866 } 4867 return false; 4868 } 4869 4870 static const char *GetOSName(uint32_t cmd) { 4871 switch (cmd) { 4872 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4873 return "ios"; 4874 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4875 return "macosx"; 4876 case llvm::MachO::LC_VERSION_MIN_TVOS: 4877 return "tvos"; 4878 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4879 return "watchos"; 4880 default: 4881 llvm_unreachable("unexpected LC_VERSION load command"); 4882 } 4883 } 4884 4885 bool ObjectFileMachO::GetArchitecture(const llvm::MachO::mach_header &header, 4886 const lldb_private::DataExtractor &data, 4887 lldb::offset_t lc_offset, 4888 ArchSpec &arch) { 4889 arch.SetArchitecture(eArchTypeMachO, header.cputype, header.cpusubtype); 4890 4891 if (arch.IsValid()) { 4892 llvm::Triple &triple = arch.GetTriple(); 4893 4894 // Set OS to an unspecified unknown or a "*" so it can match any OS 4895 triple.setOS(llvm::Triple::UnknownOS); 4896 triple.setOSName(llvm::StringRef()); 4897 4898 if (header.filetype == MH_PRELOAD) { 4899 if (header.cputype == CPU_TYPE_ARM) { 4900 // If this is a 32-bit arm binary, and it's a standalone binary, force 4901 // the Vendor to Apple so we don't accidentally pick up the generic 4902 // armv7 ABI at runtime. Apple's armv7 ABI always uses r7 for the 4903 // frame pointer register; most other armv7 ABIs use a combination of 4904 // r7 and r11. 4905 triple.setVendor(llvm::Triple::Apple); 4906 } else { 4907 // Set vendor to an unspecified unknown or a "*" so it can match any 4908 // vendor This is required for correct behavior of EFI debugging on 4909 // x86_64 4910 triple.setVendor(llvm::Triple::UnknownVendor); 4911 triple.setVendorName(llvm::StringRef()); 4912 } 4913 return true; 4914 } else { 4915 struct load_command load_cmd; 4916 4917 lldb::offset_t offset = lc_offset; 4918 for (uint32_t i = 0; i < header.ncmds; ++i) { 4919 const lldb::offset_t cmd_offset = offset; 4920 if (data.GetU32(&offset, &load_cmd, 2) == NULL) 4921 break; 4922 4923 uint32_t major, minor, patch; 4924 struct version_min_command version_min; 4925 4926 llvm::SmallString<16> os_name; 4927 llvm::raw_svector_ostream os(os_name); 4928 4929 switch (load_cmd.cmd) { 4930 case llvm::MachO::LC_VERSION_MIN_IPHONEOS: 4931 case llvm::MachO::LC_VERSION_MIN_MACOSX: 4932 case llvm::MachO::LC_VERSION_MIN_TVOS: 4933 case llvm::MachO::LC_VERSION_MIN_WATCHOS: 4934 if (load_cmd.cmdsize != sizeof(version_min)) 4935 break; 4936 data.ExtractBytes(cmd_offset, 4937 sizeof(version_min), data.GetByteOrder(), 4938 &version_min); 4939 major = version_min.version >> 16; 4940 minor = (version_min.version >> 8) & 0xffu; 4941 patch = version_min.version & 0xffu; 4942 os << GetOSName(load_cmd.cmd) << major << '.' << minor << '.' 4943 << patch; 4944 triple.setOSName(os.str()); 4945 return true; 4946 default: 4947 break; 4948 } 4949 4950 offset = cmd_offset + load_cmd.cmdsize; 4951 } 4952 4953 if (header.filetype != MH_KEXT_BUNDLE) { 4954 // We didn't find a LC_VERSION_MIN load command and this isn't a KEXT 4955 // so lets not say our Vendor is Apple, leave it as an unspecified 4956 // unknown 4957 triple.setVendor(llvm::Triple::UnknownVendor); 4958 triple.setVendorName(llvm::StringRef()); 4959 } 4960 } 4961 } 4962 return arch.IsValid(); 4963 } 4964 4965 bool ObjectFileMachO::GetUUID(lldb_private::UUID *uuid) { 4966 ModuleSP module_sp(GetModule()); 4967 if (module_sp) { 4968 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4969 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4970 return GetUUID(m_header, m_data, offset, *uuid); 4971 } 4972 return false; 4973 } 4974 4975 uint32_t ObjectFileMachO::GetDependentModules(FileSpecList &files) { 4976 uint32_t count = 0; 4977 ModuleSP module_sp(GetModule()); 4978 if (module_sp) { 4979 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 4980 struct load_command load_cmd; 4981 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 4982 std::vector<std::string> rpath_paths; 4983 std::vector<std::string> rpath_relative_paths; 4984 std::vector<std::string> at_exec_relative_paths; 4985 const bool resolve_path = false; // Don't resolve the dependent file paths 4986 // since they may not reside on this 4987 // system 4988 uint32_t i; 4989 for (i = 0; i < m_header.ncmds; ++i) { 4990 const uint32_t cmd_offset = offset; 4991 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 4992 break; 4993 4994 switch (load_cmd.cmd) { 4995 case LC_RPATH: 4996 case LC_LOAD_DYLIB: 4997 case LC_LOAD_WEAK_DYLIB: 4998 case LC_REEXPORT_DYLIB: 4999 case LC_LOAD_DYLINKER: 5000 case LC_LOADFVMLIB: 5001 case LC_LOAD_UPWARD_DYLIB: { 5002 uint32_t name_offset = cmd_offset + m_data.GetU32(&offset); 5003 const char *path = m_data.PeekCStr(name_offset); 5004 if (path) { 5005 if (load_cmd.cmd == LC_RPATH) 5006 rpath_paths.push_back(path); 5007 else { 5008 if (path[0] == '@') { 5009 if (strncmp(path, "@rpath", strlen("@rpath")) == 0) 5010 rpath_relative_paths.push_back(path + strlen("@rpath")); 5011 else if (strncmp(path, "@executable_path", 5012 strlen("@executable_path")) == 0) 5013 at_exec_relative_paths.push_back(path 5014 + strlen("@executable_path")); 5015 } else { 5016 FileSpec file_spec(path, resolve_path); 5017 if (files.AppendIfUnique(file_spec)) 5018 count++; 5019 } 5020 } 5021 } 5022 } break; 5023 5024 default: 5025 break; 5026 } 5027 offset = cmd_offset + load_cmd.cmdsize; 5028 } 5029 5030 FileSpec this_file_spec(m_file); 5031 this_file_spec.ResolvePath(); 5032 5033 if (!rpath_paths.empty()) { 5034 // Fixup all LC_RPATH values to be absolute paths 5035 std::string loader_path("@loader_path"); 5036 std::string executable_path("@executable_path"); 5037 for (auto &rpath : rpath_paths) { 5038 if (rpath.find(loader_path) == 0) { 5039 rpath.erase(0, loader_path.size()); 5040 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5041 } else if (rpath.find(executable_path) == 0) { 5042 rpath.erase(0, executable_path.size()); 5043 rpath.insert(0, this_file_spec.GetDirectory().GetCString()); 5044 } 5045 } 5046 5047 for (const auto &rpath_relative_path : rpath_relative_paths) { 5048 for (const auto &rpath : rpath_paths) { 5049 std::string path = rpath; 5050 path += rpath_relative_path; 5051 // It is OK to resolve this path because we must find a file on disk 5052 // for us to accept it anyway if it is rpath relative. 5053 FileSpec file_spec(path, true); 5054 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) { 5055 count++; 5056 break; 5057 } 5058 } 5059 } 5060 } 5061 5062 // We may have @executable_paths but no RPATHS. Figure those out here. 5063 // Only do this if this object file is the executable. We have no way to 5064 // get back to the actual executable otherwise, so we won't get the right 5065 // path. 5066 if (!at_exec_relative_paths.empty() && CalculateType() == eTypeExecutable) { 5067 FileSpec exec_dir = this_file_spec.CopyByRemovingLastPathComponent(); 5068 for (const auto &at_exec_relative_path : at_exec_relative_paths) { 5069 FileSpec file_spec = 5070 exec_dir.CopyByAppendingPathComponent(at_exec_relative_path); 5071 if (file_spec.Exists() && files.AppendIfUnique(file_spec)) 5072 count++; 5073 } 5074 } 5075 } 5076 return count; 5077 } 5078 5079 lldb_private::Address ObjectFileMachO::GetEntryPointAddress() { 5080 // If the object file is not an executable it can't hold the entry point. 5081 // m_entry_point_address is initialized to an invalid address, so we can just 5082 // return that. If m_entry_point_address is valid it means we've found it 5083 // already, so return the cached value. 5084 5085 if (!IsExecutable() || m_entry_point_address.IsValid()) 5086 return m_entry_point_address; 5087 5088 // Otherwise, look for the UnixThread or Thread command. The data for the 5089 // Thread command is given in /usr/include/mach-o.h, but it is basically: 5090 // 5091 // uint32_t flavor - this is the flavor argument you would pass to 5092 // thread_get_state 5093 // uint32_t count - this is the count of longs in the thread state data 5094 // struct XXX_thread_state state - this is the structure from 5095 // <machine/thread_status.h> corresponding to the flavor. 5096 // <repeat this trio> 5097 // 5098 // So we just keep reading the various register flavors till we find the GPR 5099 // one, then read the PC out of there. 5100 // FIXME: We will need to have a "RegisterContext data provider" class at some 5101 // point that can get all the registers 5102 // out of data in this form & attach them to a given thread. That should 5103 // underlie the MacOS X User process plugin, and we'll also need it for the 5104 // MacOS X Core File process plugin. When we have that we can also use it 5105 // here. 5106 // 5107 // For now we hard-code the offsets and flavors we need: 5108 // 5109 // 5110 5111 ModuleSP module_sp(GetModule()); 5112 if (module_sp) { 5113 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5114 struct load_command load_cmd; 5115 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5116 uint32_t i; 5117 lldb::addr_t start_address = LLDB_INVALID_ADDRESS; 5118 bool done = false; 5119 5120 for (i = 0; i < m_header.ncmds; ++i) { 5121 const lldb::offset_t cmd_offset = offset; 5122 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5123 break; 5124 5125 switch (load_cmd.cmd) { 5126 case LC_UNIXTHREAD: 5127 case LC_THREAD: { 5128 while (offset < cmd_offset + load_cmd.cmdsize) { 5129 uint32_t flavor = m_data.GetU32(&offset); 5130 uint32_t count = m_data.GetU32(&offset); 5131 if (count == 0) { 5132 // We've gotten off somehow, log and exit; 5133 return m_entry_point_address; 5134 } 5135 5136 switch (m_header.cputype) { 5137 case llvm::MachO::CPU_TYPE_ARM: 5138 if (flavor == 1 || 5139 flavor == 9) // ARM_THREAD_STATE/ARM_THREAD_STATE32 from 5140 // mach/arm/thread_status.h 5141 { 5142 offset += 60; // This is the offset of pc in the GPR thread state 5143 // data structure. 5144 start_address = m_data.GetU32(&offset); 5145 done = true; 5146 } 5147 break; 5148 case llvm::MachO::CPU_TYPE_ARM64: 5149 if (flavor == 6) // ARM_THREAD_STATE64 from mach/arm/thread_status.h 5150 { 5151 offset += 256; // This is the offset of pc in the GPR thread state 5152 // data structure. 5153 start_address = m_data.GetU64(&offset); 5154 done = true; 5155 } 5156 break; 5157 case llvm::MachO::CPU_TYPE_I386: 5158 if (flavor == 5159 1) // x86_THREAD_STATE32 from mach/i386/thread_status.h 5160 { 5161 offset += 40; // This is the offset of eip in the GPR thread state 5162 // data structure. 5163 start_address = m_data.GetU32(&offset); 5164 done = true; 5165 } 5166 break; 5167 case llvm::MachO::CPU_TYPE_X86_64: 5168 if (flavor == 5169 4) // x86_THREAD_STATE64 from mach/i386/thread_status.h 5170 { 5171 offset += 16 * 8; // This is the offset of rip in the GPR thread 5172 // state data structure. 5173 start_address = m_data.GetU64(&offset); 5174 done = true; 5175 } 5176 break; 5177 default: 5178 return m_entry_point_address; 5179 } 5180 // Haven't found the GPR flavor yet, skip over the data for this 5181 // flavor: 5182 if (done) 5183 break; 5184 offset += count * 4; 5185 } 5186 } break; 5187 case LC_MAIN: { 5188 ConstString text_segment_name("__TEXT"); 5189 uint64_t entryoffset = m_data.GetU64(&offset); 5190 SectionSP text_segment_sp = 5191 GetSectionList()->FindSectionByName(text_segment_name); 5192 if (text_segment_sp) { 5193 done = true; 5194 start_address = text_segment_sp->GetFileAddress() + entryoffset; 5195 } 5196 } break; 5197 5198 default: 5199 break; 5200 } 5201 if (done) 5202 break; 5203 5204 // Go to the next load command: 5205 offset = cmd_offset + load_cmd.cmdsize; 5206 } 5207 5208 if (start_address != LLDB_INVALID_ADDRESS) { 5209 // We got the start address from the load commands, so now resolve that 5210 // address in the sections of this ObjectFile: 5211 if (!m_entry_point_address.ResolveAddressUsingFileSections( 5212 start_address, GetSectionList())) { 5213 m_entry_point_address.Clear(); 5214 } 5215 } else { 5216 // We couldn't read the UnixThread load command - maybe it wasn't there. 5217 // As a fallback look for the "start" symbol in the main executable. 5218 5219 ModuleSP module_sp(GetModule()); 5220 5221 if (module_sp) { 5222 SymbolContextList contexts; 5223 SymbolContext context; 5224 if (module_sp->FindSymbolsWithNameAndType(ConstString("start"), 5225 eSymbolTypeCode, contexts)) { 5226 if (contexts.GetContextAtIndex(0, context)) 5227 m_entry_point_address = context.symbol->GetAddress(); 5228 } 5229 } 5230 } 5231 } 5232 5233 return m_entry_point_address; 5234 } 5235 5236 lldb_private::Address ObjectFileMachO::GetHeaderAddress() { 5237 lldb_private::Address header_addr; 5238 SectionList *section_list = GetSectionList(); 5239 if (section_list) { 5240 SectionSP text_segment_sp( 5241 section_list->FindSectionByName(GetSegmentNameTEXT())); 5242 if (text_segment_sp) { 5243 header_addr.SetSection(text_segment_sp); 5244 header_addr.SetOffset(0); 5245 } 5246 } 5247 return header_addr; 5248 } 5249 5250 uint32_t ObjectFileMachO::GetNumThreadContexts() { 5251 ModuleSP module_sp(GetModule()); 5252 if (module_sp) { 5253 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5254 if (!m_thread_context_offsets_valid) { 5255 m_thread_context_offsets_valid = true; 5256 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5257 FileRangeArray::Entry file_range; 5258 thread_command thread_cmd; 5259 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5260 const uint32_t cmd_offset = offset; 5261 if (m_data.GetU32(&offset, &thread_cmd, 2) == NULL) 5262 break; 5263 5264 if (thread_cmd.cmd == LC_THREAD) { 5265 file_range.SetRangeBase(offset); 5266 file_range.SetByteSize(thread_cmd.cmdsize - 8); 5267 m_thread_context_offsets.Append(file_range); 5268 } 5269 offset = cmd_offset + thread_cmd.cmdsize; 5270 } 5271 } 5272 } 5273 return m_thread_context_offsets.GetSize(); 5274 } 5275 5276 std::string ObjectFileMachO::GetIdentifierString() { 5277 std::string result; 5278 ModuleSP module_sp(GetModule()); 5279 if (module_sp) { 5280 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5281 5282 // First, look over the load commands for an LC_NOTE load command with 5283 // data_owner string "kern ver str" & use that if found. 5284 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5285 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5286 const uint32_t cmd_offset = offset; 5287 load_command lc; 5288 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5289 break; 5290 if (lc.cmd == LC_NOTE) 5291 { 5292 char data_owner[17]; 5293 m_data.CopyData (offset, 16, data_owner); 5294 data_owner[16] = '\0'; 5295 offset += 16; 5296 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5297 uint64_t size = m_data.GetU64_unchecked (&offset); 5298 5299 // "kern ver str" has a uint32_t version and then a nul terminated 5300 // c-string. 5301 if (strcmp ("kern ver str", data_owner) == 0) 5302 { 5303 offset = fileoff; 5304 uint32_t version; 5305 if (m_data.GetU32 (&offset, &version, 1) != nullptr) 5306 { 5307 if (version == 1) 5308 { 5309 uint32_t strsize = size - sizeof (uint32_t); 5310 char *buf = (char*) malloc (strsize); 5311 if (buf) 5312 { 5313 m_data.CopyData (offset, strsize, buf); 5314 buf[strsize - 1] = '\0'; 5315 result = buf; 5316 if (buf) 5317 free (buf); 5318 return result; 5319 } 5320 } 5321 } 5322 } 5323 } 5324 offset = cmd_offset + lc.cmdsize; 5325 } 5326 5327 // Second, make a pass over the load commands looking for an obsolete 5328 // LC_IDENT load command. 5329 offset = MachHeaderSizeFromMagic(m_header.magic); 5330 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5331 const uint32_t cmd_offset = offset; 5332 struct ident_command ident_command; 5333 if (m_data.GetU32(&offset, &ident_command, 2) == NULL) 5334 break; 5335 if (ident_command.cmd == LC_IDENT && ident_command.cmdsize != 0) { 5336 char *buf = (char *) malloc (ident_command.cmdsize); 5337 if (buf != nullptr 5338 && m_data.CopyData (offset, ident_command.cmdsize, buf) == ident_command.cmdsize) { 5339 buf[ident_command.cmdsize - 1] = '\0'; 5340 result = buf; 5341 } 5342 if (buf) 5343 free (buf); 5344 } 5345 offset = cmd_offset + ident_command.cmdsize; 5346 } 5347 5348 } 5349 return result; 5350 } 5351 5352 bool ObjectFileMachO::GetCorefileMainBinaryInfo (addr_t &address, UUID &uuid) { 5353 address = LLDB_INVALID_ADDRESS; 5354 uuid.Clear(); 5355 ModuleSP module_sp(GetModule()); 5356 if (module_sp) { 5357 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5358 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5359 for (uint32_t i = 0; i < m_header.ncmds; ++i) { 5360 const uint32_t cmd_offset = offset; 5361 load_command lc; 5362 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5363 break; 5364 if (lc.cmd == LC_NOTE) 5365 { 5366 char data_owner[17]; 5367 memset (data_owner, 0, sizeof (data_owner)); 5368 m_data.CopyData (offset, 16, data_owner); 5369 offset += 16; 5370 uint64_t fileoff = m_data.GetU64_unchecked (&offset); 5371 uint64_t size = m_data.GetU64_unchecked (&offset); 5372 5373 // "main bin spec" (main binary specification) data payload is 5374 // formatted: 5375 // uint32_t version [currently 1] 5376 // uint32_t type [0 == unspecified, 1 == kernel, 2 == user process] 5377 // uint64_t address [ UINT64_MAX if address not specified ] 5378 // uuid_t uuid [ all zero's if uuid not specified ] 5379 // uint32_t log2_pagesize [ process page size in log base 2, e.g. 4k pages are 12. 0 for unspecified ] 5380 5381 if (strcmp ("main bin spec", data_owner) == 0 && size >= 32) 5382 { 5383 offset = fileoff; 5384 uint32_t version; 5385 if (m_data.GetU32 (&offset, &version, 1) != nullptr && version == 1) 5386 { 5387 uint32_t type = 0; 5388 uuid_t raw_uuid; 5389 memset (raw_uuid, 0, sizeof (uuid_t)); 5390 5391 if (m_data.GetU32 (&offset, &type, 1) 5392 && m_data.GetU64 (&offset, &address, 1) 5393 && m_data.CopyData (offset, sizeof (uuid_t), raw_uuid) != 0 5394 && uuid.SetBytes (raw_uuid, sizeof (uuid_t))) 5395 { 5396 return true; 5397 } 5398 } 5399 } 5400 } 5401 offset = cmd_offset + lc.cmdsize; 5402 } 5403 } 5404 return false; 5405 } 5406 5407 lldb::RegisterContextSP 5408 ObjectFileMachO::GetThreadContextAtIndex(uint32_t idx, 5409 lldb_private::Thread &thread) { 5410 lldb::RegisterContextSP reg_ctx_sp; 5411 5412 ModuleSP module_sp(GetModule()); 5413 if (module_sp) { 5414 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5415 if (!m_thread_context_offsets_valid) 5416 GetNumThreadContexts(); 5417 5418 const FileRangeArray::Entry *thread_context_file_range = 5419 m_thread_context_offsets.GetEntryAtIndex(idx); 5420 if (thread_context_file_range) { 5421 5422 DataExtractor data(m_data, thread_context_file_range->GetRangeBase(), 5423 thread_context_file_range->GetByteSize()); 5424 5425 switch (m_header.cputype) { 5426 case llvm::MachO::CPU_TYPE_ARM64: 5427 reg_ctx_sp.reset(new RegisterContextDarwin_arm64_Mach(thread, data)); 5428 break; 5429 5430 case llvm::MachO::CPU_TYPE_ARM: 5431 reg_ctx_sp.reset(new RegisterContextDarwin_arm_Mach(thread, data)); 5432 break; 5433 5434 case llvm::MachO::CPU_TYPE_I386: 5435 reg_ctx_sp.reset(new RegisterContextDarwin_i386_Mach(thread, data)); 5436 break; 5437 5438 case llvm::MachO::CPU_TYPE_X86_64: 5439 reg_ctx_sp.reset(new RegisterContextDarwin_x86_64_Mach(thread, data)); 5440 break; 5441 } 5442 } 5443 } 5444 return reg_ctx_sp; 5445 } 5446 5447 ObjectFile::Type ObjectFileMachO::CalculateType() { 5448 switch (m_header.filetype) { 5449 case MH_OBJECT: // 0x1u 5450 if (GetAddressByteSize() == 4) { 5451 // 32 bit kexts are just object files, but they do have a valid 5452 // UUID load command. 5453 UUID uuid; 5454 if (GetUUID(&uuid)) { 5455 // this checking for the UUID load command is not enough we could 5456 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5457 // this is required of kexts 5458 if (m_strata == eStrataInvalid) 5459 m_strata = eStrataKernel; 5460 return eTypeSharedLibrary; 5461 } 5462 } 5463 return eTypeObjectFile; 5464 5465 case MH_EXECUTE: 5466 return eTypeExecutable; // 0x2u 5467 case MH_FVMLIB: 5468 return eTypeSharedLibrary; // 0x3u 5469 case MH_CORE: 5470 return eTypeCoreFile; // 0x4u 5471 case MH_PRELOAD: 5472 return eTypeSharedLibrary; // 0x5u 5473 case MH_DYLIB: 5474 return eTypeSharedLibrary; // 0x6u 5475 case MH_DYLINKER: 5476 return eTypeDynamicLinker; // 0x7u 5477 case MH_BUNDLE: 5478 return eTypeSharedLibrary; // 0x8u 5479 case MH_DYLIB_STUB: 5480 return eTypeStubLibrary; // 0x9u 5481 case MH_DSYM: 5482 return eTypeDebugInfo; // 0xAu 5483 case MH_KEXT_BUNDLE: 5484 return eTypeSharedLibrary; // 0xBu 5485 default: 5486 break; 5487 } 5488 return eTypeUnknown; 5489 } 5490 5491 ObjectFile::Strata ObjectFileMachO::CalculateStrata() { 5492 switch (m_header.filetype) { 5493 case MH_OBJECT: // 0x1u 5494 { 5495 // 32 bit kexts are just object files, but they do have a valid 5496 // UUID load command. 5497 UUID uuid; 5498 if (GetUUID(&uuid)) { 5499 // this checking for the UUID load command is not enough we could 5500 // eventually look for the symbol named "OSKextGetCurrentIdentifier" as 5501 // this is required of kexts 5502 if (m_type == eTypeInvalid) 5503 m_type = eTypeSharedLibrary; 5504 5505 return eStrataKernel; 5506 } 5507 } 5508 return eStrataUnknown; 5509 5510 case MH_EXECUTE: // 0x2u 5511 // Check for the MH_DYLDLINK bit in the flags 5512 if (m_header.flags & MH_DYLDLINK) { 5513 return eStrataUser; 5514 } else { 5515 SectionList *section_list = GetSectionList(); 5516 if (section_list) { 5517 static ConstString g_kld_section_name("__KLD"); 5518 if (section_list->FindSectionByName(g_kld_section_name)) 5519 return eStrataKernel; 5520 } 5521 } 5522 return eStrataRawImage; 5523 5524 case MH_FVMLIB: 5525 return eStrataUser; // 0x3u 5526 case MH_CORE: 5527 return eStrataUnknown; // 0x4u 5528 case MH_PRELOAD: 5529 return eStrataRawImage; // 0x5u 5530 case MH_DYLIB: 5531 return eStrataUser; // 0x6u 5532 case MH_DYLINKER: 5533 return eStrataUser; // 0x7u 5534 case MH_BUNDLE: 5535 return eStrataUser; // 0x8u 5536 case MH_DYLIB_STUB: 5537 return eStrataUser; // 0x9u 5538 case MH_DSYM: 5539 return eStrataUnknown; // 0xAu 5540 case MH_KEXT_BUNDLE: 5541 return eStrataKernel; // 0xBu 5542 default: 5543 break; 5544 } 5545 return eStrataUnknown; 5546 } 5547 5548 uint32_t ObjectFileMachO::GetVersion(uint32_t *versions, 5549 uint32_t num_versions) { 5550 ModuleSP module_sp(GetModule()); 5551 if (module_sp) { 5552 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5553 struct dylib_command load_cmd; 5554 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5555 uint32_t version_cmd = 0; 5556 uint64_t version = 0; 5557 uint32_t i; 5558 for (i = 0; i < m_header.ncmds; ++i) { 5559 const lldb::offset_t cmd_offset = offset; 5560 if (m_data.GetU32(&offset, &load_cmd, 2) == NULL) 5561 break; 5562 5563 if (load_cmd.cmd == LC_ID_DYLIB) { 5564 if (version_cmd == 0) { 5565 version_cmd = load_cmd.cmd; 5566 if (m_data.GetU32(&offset, &load_cmd.dylib, 4) == NULL) 5567 break; 5568 version = load_cmd.dylib.current_version; 5569 } 5570 break; // Break for now unless there is another more complete version 5571 // number load command in the future. 5572 } 5573 offset = cmd_offset + load_cmd.cmdsize; 5574 } 5575 5576 if (version_cmd == LC_ID_DYLIB) { 5577 if (versions != NULL && num_versions > 0) { 5578 if (num_versions > 0) 5579 versions[0] = (version & 0xFFFF0000ull) >> 16; 5580 if (num_versions > 1) 5581 versions[1] = (version & 0x0000FF00ull) >> 8; 5582 if (num_versions > 2) 5583 versions[2] = (version & 0x000000FFull); 5584 // Fill in an remaining version numbers with invalid values 5585 for (i = 3; i < num_versions; ++i) 5586 versions[i] = UINT32_MAX; 5587 } 5588 // The LC_ID_DYLIB load command has a version with 3 version numbers in 5589 // it, so always return 3 5590 return 3; 5591 } 5592 } 5593 return false; 5594 } 5595 5596 bool ObjectFileMachO::GetArchitecture(ArchSpec &arch) { 5597 ModuleSP module_sp(GetModule()); 5598 if (module_sp) { 5599 std::lock_guard<std::recursive_mutex> guard(module_sp->GetMutex()); 5600 return GetArchitecture(m_header, m_data, 5601 MachHeaderSizeFromMagic(m_header.magic), arch); 5602 } 5603 return false; 5604 } 5605 5606 void ObjectFileMachO::GetProcessSharedCacheUUID(Process *process, addr_t &base_addr, UUID &uuid) { 5607 uuid.Clear(); 5608 base_addr = LLDB_INVALID_ADDRESS; 5609 if (process && process->GetDynamicLoader()) { 5610 DynamicLoader *dl = process->GetDynamicLoader(); 5611 LazyBool using_shared_cache; 5612 LazyBool private_shared_cache; 5613 dl->GetSharedCacheInformation(base_addr, uuid, using_shared_cache, 5614 private_shared_cache); 5615 } 5616 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5617 if (log) 5618 log->Printf("inferior process shared cache has a UUID of %s, base address 0x%" PRIx64 , uuid.GetAsString().c_str(), base_addr); 5619 } 5620 5621 // From dyld SPI header dyld_process_info.h 5622 typedef void *dyld_process_info; 5623 struct lldb_copy__dyld_process_cache_info { 5624 uuid_t cacheUUID; // UUID of cache used by process 5625 uint64_t cacheBaseAddress; // load address of dyld shared cache 5626 bool noCache; // process is running without a dyld cache 5627 bool privateCache; // process is using a private copy of its dyld cache 5628 }; 5629 5630 // #including mach/mach.h pulls in machine.h & CPU_TYPE_ARM etc conflicts with llvm 5631 // enum definitions llvm::MachO::CPU_TYPE_ARM turning them into compile errors. 5632 // So we need to use the actual underlying types of task_t and kern_return_t 5633 // below. 5634 extern "C" unsigned int /*task_t*/ mach_task_self(); 5635 5636 void ObjectFileMachO::GetLLDBSharedCacheUUID(addr_t &base_addr, UUID &uuid) { 5637 uuid.Clear(); 5638 base_addr = LLDB_INVALID_ADDRESS; 5639 5640 #if defined(__APPLE__) && \ 5641 (defined(__arm__) || defined(__arm64__) || defined(__aarch64__)) 5642 uint8_t *(*dyld_get_all_image_infos)(void); 5643 dyld_get_all_image_infos = 5644 (uint8_t * (*)())dlsym(RTLD_DEFAULT, "_dyld_get_all_image_infos"); 5645 if (dyld_get_all_image_infos) { 5646 uint8_t *dyld_all_image_infos_address = dyld_get_all_image_infos(); 5647 if (dyld_all_image_infos_address) { 5648 uint32_t *version = (uint32_t *) 5649 dyld_all_image_infos_address; // version <mach-o/dyld_images.h> 5650 if (*version >= 13) { 5651 uuid_t *sharedCacheUUID_address = 0; 5652 int wordsize = sizeof(uint8_t *); 5653 if (wordsize == 8) { 5654 sharedCacheUUID_address = 5655 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5656 160); // sharedCacheUUID <mach-o/dyld_images.h> 5657 if (*version >= 15) 5658 base_addr = *(uint64_t *) ((uint8_t *) dyld_all_image_infos_address 5659 + 176); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5660 } else { 5661 sharedCacheUUID_address = 5662 (uuid_t *)((uint8_t *)dyld_all_image_infos_address + 5663 84); // sharedCacheUUID <mach-o/dyld_images.h> 5664 if (*version >= 15) { 5665 base_addr = 0; 5666 base_addr = *(uint32_t *) ((uint8_t *) dyld_all_image_infos_address 5667 + 100); // sharedCacheBaseAddress <mach-o/dyld_images.h> 5668 } 5669 } 5670 uuid.SetBytes(sharedCacheUUID_address); 5671 } 5672 } 5673 } else { 5674 // Exists in macOS 10.12 and later, iOS 10.0 and later - dyld SPI 5675 dyld_process_info (*dyld_process_info_create)(unsigned int /* task_t */ task, uint64_t timestamp, unsigned int /*kern_return_t*/ *kernelError); 5676 void (*dyld_process_info_get_cache)(void *info, void *cacheInfo); 5677 void (*dyld_process_info_release)(dyld_process_info info); 5678 5679 dyld_process_info_create = (void *(*)(unsigned int /* task_t */, uint64_t, unsigned int /*kern_return_t*/ *)) 5680 dlsym (RTLD_DEFAULT, "_dyld_process_info_create"); 5681 dyld_process_info_get_cache = (void (*)(void *, void *)) 5682 dlsym (RTLD_DEFAULT, "_dyld_process_info_get_cache"); 5683 dyld_process_info_release = (void (*)(void *)) 5684 dlsym (RTLD_DEFAULT, "_dyld_process_info_release"); 5685 5686 if (dyld_process_info_create && dyld_process_info_get_cache) { 5687 unsigned int /*kern_return_t */ kern_ret; 5688 dyld_process_info process_info = dyld_process_info_create(::mach_task_self(), 0, &kern_ret); 5689 if (process_info) { 5690 struct lldb_copy__dyld_process_cache_info sc_info; 5691 memset (&sc_info, 0, sizeof (struct lldb_copy__dyld_process_cache_info)); 5692 dyld_process_info_get_cache (process_info, &sc_info); 5693 if (sc_info.cacheBaseAddress != 0) { 5694 base_addr = sc_info.cacheBaseAddress; 5695 uuid.SetBytes (sc_info.cacheUUID); 5696 } 5697 dyld_process_info_release (process_info); 5698 } 5699 } 5700 } 5701 Log *log(lldb_private::GetLogIfAnyCategoriesSet(LIBLLDB_LOG_SYMBOLS | LIBLLDB_LOG_PROCESS)); 5702 if (log && uuid.IsValid()) 5703 log->Printf("lldb's in-memory shared cache has a UUID of %s base address of 0x%" PRIx64, uuid.GetAsString().c_str(), base_addr); 5704 #endif 5705 } 5706 5707 uint32_t ObjectFileMachO::GetMinimumOSVersion(uint32_t *versions, 5708 uint32_t num_versions) { 5709 if (m_min_os_versions.empty()) { 5710 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5711 bool success = false; 5712 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5713 const lldb::offset_t load_cmd_offset = offset; 5714 5715 version_min_command lc; 5716 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5717 break; 5718 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5719 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5720 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5721 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5722 if (m_data.GetU32(&offset, &lc.version, 5723 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5724 const uint32_t xxxx = lc.version >> 16; 5725 const uint32_t yy = (lc.version >> 8) & 0xffu; 5726 const uint32_t zz = lc.version & 0xffu; 5727 if (xxxx) { 5728 m_min_os_versions.push_back(xxxx); 5729 m_min_os_versions.push_back(yy); 5730 m_min_os_versions.push_back(zz); 5731 } 5732 success = true; 5733 } 5734 } 5735 offset = load_cmd_offset + lc.cmdsize; 5736 } 5737 5738 if (success == false) { 5739 // Push an invalid value so we don't keep trying to 5740 m_min_os_versions.push_back(UINT32_MAX); 5741 } 5742 } 5743 5744 if (m_min_os_versions.size() > 1 || m_min_os_versions[0] != UINT32_MAX) { 5745 if (versions != NULL && num_versions > 0) { 5746 for (size_t i = 0; i < num_versions; ++i) { 5747 if (i < m_min_os_versions.size()) 5748 versions[i] = m_min_os_versions[i]; 5749 else 5750 versions[i] = 0; 5751 } 5752 } 5753 return m_min_os_versions.size(); 5754 } 5755 // Call the superclasses version that will empty out the data 5756 return ObjectFile::GetMinimumOSVersion(versions, num_versions); 5757 } 5758 5759 uint32_t ObjectFileMachO::GetSDKVersion(uint32_t *versions, 5760 uint32_t num_versions) { 5761 if (m_sdk_versions.empty()) { 5762 lldb::offset_t offset = MachHeaderSizeFromMagic(m_header.magic); 5763 bool success = false; 5764 for (uint32_t i = 0; success == false && i < m_header.ncmds; ++i) { 5765 const lldb::offset_t load_cmd_offset = offset; 5766 5767 version_min_command lc; 5768 if (m_data.GetU32(&offset, &lc.cmd, 2) == NULL) 5769 break; 5770 if (lc.cmd == llvm::MachO::LC_VERSION_MIN_MACOSX || 5771 lc.cmd == llvm::MachO::LC_VERSION_MIN_IPHONEOS || 5772 lc.cmd == llvm::MachO::LC_VERSION_MIN_TVOS || 5773 lc.cmd == llvm::MachO::LC_VERSION_MIN_WATCHOS) { 5774 if (m_data.GetU32(&offset, &lc.version, 5775 (sizeof(lc) / sizeof(uint32_t)) - 2)) { 5776 const uint32_t xxxx = lc.sdk >> 16; 5777 const uint32_t yy = (lc.sdk >> 8) & 0xffu; 5778 const uint32_t zz = lc.sdk & 0xffu; 5779 if (xxxx) { 5780 m_sdk_versions.push_back(xxxx); 5781 m_sdk_versions.push_back(yy); 5782 m_sdk_versions.push_back(zz); 5783 } 5784 success = true; 5785 } 5786 } 5787 offset = load_cmd_offset + lc.cmdsize; 5788 } 5789 5790 if (success == false) { 5791 // Push an invalid value so we don't keep trying to 5792 m_sdk_versions.push_back(UINT32_MAX); 5793 } 5794 } 5795 5796 if (m_sdk_versions.size() > 1 || m_sdk_versions[0] != UINT32_MAX) { 5797 if (versions != NULL && num_versions > 0) { 5798 for (size_t i = 0; i < num_versions; ++i) { 5799 if (i < m_sdk_versions.size()) 5800 versions[i] = m_sdk_versions[i]; 5801 else 5802 versions[i] = 0; 5803 } 5804 } 5805 return m_sdk_versions.size(); 5806 } 5807 // Call the superclasses version that will empty out the data 5808 return ObjectFile::GetSDKVersion(versions, num_versions); 5809 } 5810 5811 bool ObjectFileMachO::GetIsDynamicLinkEditor() { 5812 return m_header.filetype == llvm::MachO::MH_DYLINKER; 5813 } 5814 5815 bool ObjectFileMachO::AllowAssemblyEmulationUnwindPlans() { 5816 return m_allow_assembly_emulation_unwind_plans; 5817 } 5818 5819 //------------------------------------------------------------------ 5820 // PluginInterface protocol 5821 //------------------------------------------------------------------ 5822 lldb_private::ConstString ObjectFileMachO::GetPluginName() { 5823 return GetPluginNameStatic(); 5824 } 5825 5826 uint32_t ObjectFileMachO::GetPluginVersion() { return 1; } 5827 5828 Section *ObjectFileMachO::GetMachHeaderSection() { 5829 // Find the first address of the mach header which is the first non-zero file 5830 // sized section whose file offset is zero. This is the base file address of 5831 // the mach-o file which can be subtracted from the vmaddr of the other 5832 // segments found in memory and added to the load address 5833 ModuleSP module_sp = GetModule(); 5834 if (module_sp) { 5835 SectionList *section_list = GetSectionList(); 5836 if (section_list) { 5837 lldb::addr_t mach_base_file_addr = LLDB_INVALID_ADDRESS; 5838 const size_t num_sections = section_list->GetSize(); 5839 5840 for (size_t sect_idx = 0; sect_idx < num_sections && 5841 mach_base_file_addr == LLDB_INVALID_ADDRESS; 5842 ++sect_idx) { 5843 Section *section = section_list->GetSectionAtIndex(sect_idx).get(); 5844 if (section && section->GetFileSize() > 0 && 5845 section->GetFileOffset() == 0 && 5846 section->IsThreadSpecific() == false && 5847 module_sp.get() == section->GetModule().get()) { 5848 return section; 5849 } 5850 } 5851 } 5852 } 5853 return nullptr; 5854 } 5855 5856 lldb::addr_t ObjectFileMachO::CalculateSectionLoadAddressForMemoryImage( 5857 lldb::addr_t mach_header_load_address, const Section *mach_header_section, 5858 const Section *section) { 5859 ModuleSP module_sp = GetModule(); 5860 if (module_sp && mach_header_section && section && 5861 mach_header_load_address != LLDB_INVALID_ADDRESS) { 5862 lldb::addr_t mach_header_file_addr = mach_header_section->GetFileAddress(); 5863 if (mach_header_file_addr != LLDB_INVALID_ADDRESS) { 5864 if (section && section->GetFileSize() > 0 && 5865 section->IsThreadSpecific() == false && 5866 module_sp.get() == section->GetModule().get()) { 5867 // Ignore __LINKEDIT and __DWARF segments 5868 if (section->GetName() == GetSegmentNameLINKEDIT()) { 5869 // Only map __LINKEDIT if we have an in memory image and this isn't a 5870 // kernel binary like a kext or mach_kernel. 5871 const bool is_memory_image = (bool)m_process_wp.lock(); 5872 const Strata strata = GetStrata(); 5873 if (is_memory_image == false || strata == eStrataKernel) 5874 return LLDB_INVALID_ADDRESS; 5875 } 5876 return section->GetFileAddress() - mach_header_file_addr + 5877 mach_header_load_address; 5878 } 5879 } 5880 } 5881 return LLDB_INVALID_ADDRESS; 5882 } 5883 5884 bool ObjectFileMachO::SetLoadAddress(Target &target, lldb::addr_t value, 5885 bool value_is_offset) { 5886 ModuleSP module_sp = GetModule(); 5887 if (module_sp) { 5888 size_t num_loaded_sections = 0; 5889 SectionList *section_list = GetSectionList(); 5890 if (section_list) { 5891 const size_t num_sections = section_list->GetSize(); 5892 5893 if (value_is_offset) { 5894 // "value" is an offset to apply to each top level segment 5895 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5896 // Iterate through the object file sections to find all of the 5897 // sections that size on disk (to avoid __PAGEZERO) and load them 5898 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5899 if (section_sp && section_sp->GetFileSize() > 0 && 5900 section_sp->IsThreadSpecific() == false && 5901 module_sp.get() == section_sp->GetModule().get()) { 5902 // Ignore __LINKEDIT and __DWARF segments 5903 if (section_sp->GetName() == GetSegmentNameLINKEDIT()) { 5904 // Only map __LINKEDIT if we have an in memory image and this 5905 // isn't a kernel binary like a kext or mach_kernel. 5906 const bool is_memory_image = (bool)m_process_wp.lock(); 5907 const Strata strata = GetStrata(); 5908 if (is_memory_image == false || strata == eStrataKernel) 5909 continue; 5910 } 5911 if (target.GetSectionLoadList().SetSectionLoadAddress( 5912 section_sp, section_sp->GetFileAddress() + value)) 5913 ++num_loaded_sections; 5914 } 5915 } 5916 } else { 5917 // "value" is the new base address of the mach_header, adjust each 5918 // section accordingly 5919 5920 Section *mach_header_section = GetMachHeaderSection(); 5921 if (mach_header_section) { 5922 for (size_t sect_idx = 0; sect_idx < num_sections; ++sect_idx) { 5923 SectionSP section_sp(section_list->GetSectionAtIndex(sect_idx)); 5924 5925 lldb::addr_t section_load_addr = 5926 CalculateSectionLoadAddressForMemoryImage( 5927 value, mach_header_section, section_sp.get()); 5928 if (section_load_addr != LLDB_INVALID_ADDRESS) { 5929 if (target.GetSectionLoadList().SetSectionLoadAddress( 5930 section_sp, section_load_addr)) 5931 ++num_loaded_sections; 5932 } 5933 } 5934 } 5935 } 5936 } 5937 return num_loaded_sections > 0; 5938 } 5939 return false; 5940 } 5941 5942 bool ObjectFileMachO::SaveCore(const lldb::ProcessSP &process_sp, 5943 const FileSpec &outfile, Status &error) { 5944 if (process_sp) { 5945 Target &target = process_sp->GetTarget(); 5946 const ArchSpec target_arch = target.GetArchitecture(); 5947 const llvm::Triple &target_triple = target_arch.GetTriple(); 5948 if (target_triple.getVendor() == llvm::Triple::Apple && 5949 (target_triple.getOS() == llvm::Triple::MacOSX || 5950 target_triple.getOS() == llvm::Triple::IOS || 5951 target_triple.getOS() == llvm::Triple::WatchOS || 5952 target_triple.getOS() == llvm::Triple::TvOS)) { 5953 bool make_core = false; 5954 switch (target_arch.GetMachine()) { 5955 case llvm::Triple::aarch64: 5956 case llvm::Triple::arm: 5957 case llvm::Triple::thumb: 5958 case llvm::Triple::x86: 5959 case llvm::Triple::x86_64: 5960 make_core = true; 5961 break; 5962 default: 5963 error.SetErrorStringWithFormat("unsupported core architecture: %s", 5964 target_triple.str().c_str()); 5965 break; 5966 } 5967 5968 if (make_core) { 5969 std::vector<segment_command_64> segment_load_commands; 5970 // uint32_t range_info_idx = 0; 5971 MemoryRegionInfo range_info; 5972 Status range_error = process_sp->GetMemoryRegionInfo(0, range_info); 5973 const uint32_t addr_byte_size = target_arch.GetAddressByteSize(); 5974 const ByteOrder byte_order = target_arch.GetByteOrder(); 5975 if (range_error.Success()) { 5976 while (range_info.GetRange().GetRangeBase() != LLDB_INVALID_ADDRESS) { 5977 const addr_t addr = range_info.GetRange().GetRangeBase(); 5978 const addr_t size = range_info.GetRange().GetByteSize(); 5979 5980 if (size == 0) 5981 break; 5982 5983 // Calculate correct protections 5984 uint32_t prot = 0; 5985 if (range_info.GetReadable() == MemoryRegionInfo::eYes) 5986 prot |= VM_PROT_READ; 5987 if (range_info.GetWritable() == MemoryRegionInfo::eYes) 5988 prot |= VM_PROT_WRITE; 5989 if (range_info.GetExecutable() == MemoryRegionInfo::eYes) 5990 prot |= VM_PROT_EXECUTE; 5991 5992 // printf ("[%3u] [0x%16.16" PRIx64 " - 5993 // 0x%16.16" PRIx64 ") %c%c%c\n", 5994 // range_info_idx, 5995 // addr, 5996 // size, 5997 // (prot & VM_PROT_READ ) ? 'r' : 5998 // '-', 5999 // (prot & VM_PROT_WRITE ) ? 'w' : 6000 // '-', 6001 // (prot & VM_PROT_EXECUTE) ? 'x' : 6002 // '-'); 6003 6004 if (prot != 0) { 6005 uint32_t cmd_type = LC_SEGMENT_64; 6006 uint32_t segment_size = sizeof(segment_command_64); 6007 if (addr_byte_size == 4) { 6008 cmd_type = LC_SEGMENT; 6009 segment_size = sizeof(segment_command); 6010 } 6011 segment_command_64 segment = { 6012 cmd_type, // uint32_t cmd; 6013 segment_size, // uint32_t cmdsize; 6014 {0}, // char segname[16]; 6015 addr, // uint64_t vmaddr; // uint32_t for 32-bit Mach-O 6016 size, // uint64_t vmsize; // uint32_t for 32-bit Mach-O 6017 0, // uint64_t fileoff; // uint32_t for 32-bit Mach-O 6018 size, // uint64_t filesize; // uint32_t for 32-bit Mach-O 6019 prot, // uint32_t maxprot; 6020 prot, // uint32_t initprot; 6021 0, // uint32_t nsects; 6022 0}; // uint32_t flags; 6023 segment_load_commands.push_back(segment); 6024 } else { 6025 // No protections and a size of 1 used to be returned from old 6026 // debugservers when we asked about a region that was past the 6027 // last memory region and it indicates the end... 6028 if (size == 1) 6029 break; 6030 } 6031 6032 range_error = process_sp->GetMemoryRegionInfo( 6033 range_info.GetRange().GetRangeEnd(), range_info); 6034 if (range_error.Fail()) 6035 break; 6036 } 6037 6038 StreamString buffer(Stream::eBinary, addr_byte_size, byte_order); 6039 6040 mach_header_64 mach_header; 6041 if (addr_byte_size == 8) { 6042 mach_header.magic = MH_MAGIC_64; 6043 } else { 6044 mach_header.magic = MH_MAGIC; 6045 } 6046 mach_header.cputype = target_arch.GetMachOCPUType(); 6047 mach_header.cpusubtype = target_arch.GetMachOCPUSubType(); 6048 mach_header.filetype = MH_CORE; 6049 mach_header.ncmds = segment_load_commands.size(); 6050 mach_header.flags = 0; 6051 mach_header.reserved = 0; 6052 ThreadList &thread_list = process_sp->GetThreadList(); 6053 const uint32_t num_threads = thread_list.GetSize(); 6054 6055 // Make an array of LC_THREAD data items. Each one contains the 6056 // contents of the LC_THREAD load command. The data doesn't contain 6057 // the load command + load command size, we will add the load command 6058 // and load command size as we emit the data. 6059 std::vector<StreamString> LC_THREAD_datas(num_threads); 6060 for (auto &LC_THREAD_data : LC_THREAD_datas) { 6061 LC_THREAD_data.GetFlags().Set(Stream::eBinary); 6062 LC_THREAD_data.SetAddressByteSize(addr_byte_size); 6063 LC_THREAD_data.SetByteOrder(byte_order); 6064 } 6065 for (uint32_t thread_idx = 0; thread_idx < num_threads; 6066 ++thread_idx) { 6067 ThreadSP thread_sp(thread_list.GetThreadAtIndex(thread_idx)); 6068 if (thread_sp) { 6069 switch (mach_header.cputype) { 6070 case llvm::MachO::CPU_TYPE_ARM64: 6071 RegisterContextDarwin_arm64_Mach::Create_LC_THREAD( 6072 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6073 break; 6074 6075 case llvm::MachO::CPU_TYPE_ARM: 6076 RegisterContextDarwin_arm_Mach::Create_LC_THREAD( 6077 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6078 break; 6079 6080 case llvm::MachO::CPU_TYPE_I386: 6081 RegisterContextDarwin_i386_Mach::Create_LC_THREAD( 6082 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6083 break; 6084 6085 case llvm::MachO::CPU_TYPE_X86_64: 6086 RegisterContextDarwin_x86_64_Mach::Create_LC_THREAD( 6087 thread_sp.get(), LC_THREAD_datas[thread_idx]); 6088 break; 6089 } 6090 } 6091 } 6092 6093 // The size of the load command is the size of the segments... 6094 if (addr_byte_size == 8) { 6095 mach_header.sizeofcmds = segment_load_commands.size() * 6096 sizeof(struct segment_command_64); 6097 } else { 6098 mach_header.sizeofcmds = 6099 segment_load_commands.size() * sizeof(struct segment_command); 6100 } 6101 6102 // and the size of all LC_THREAD load command 6103 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6104 ++mach_header.ncmds; 6105 mach_header.sizeofcmds += 8 + LC_THREAD_data.GetSize(); 6106 } 6107 6108 printf("mach_header: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x " 6109 "0x%8.8x 0x%8.8x\n", 6110 mach_header.magic, mach_header.cputype, mach_header.cpusubtype, 6111 mach_header.filetype, mach_header.ncmds, 6112 mach_header.sizeofcmds, mach_header.flags, 6113 mach_header.reserved); 6114 6115 // Write the mach header 6116 buffer.PutHex32(mach_header.magic); 6117 buffer.PutHex32(mach_header.cputype); 6118 buffer.PutHex32(mach_header.cpusubtype); 6119 buffer.PutHex32(mach_header.filetype); 6120 buffer.PutHex32(mach_header.ncmds); 6121 buffer.PutHex32(mach_header.sizeofcmds); 6122 buffer.PutHex32(mach_header.flags); 6123 if (addr_byte_size == 8) { 6124 buffer.PutHex32(mach_header.reserved); 6125 } 6126 6127 // Skip the mach header and all load commands and align to the next 6128 // 0x1000 byte boundary 6129 addr_t file_offset = buffer.GetSize() + mach_header.sizeofcmds; 6130 if (file_offset & 0x00000fff) { 6131 file_offset += 0x00001000ull; 6132 file_offset &= (~0x00001000ull + 1); 6133 } 6134 6135 for (auto &segment : segment_load_commands) { 6136 segment.fileoff = file_offset; 6137 file_offset += segment.filesize; 6138 } 6139 6140 // Write out all of the LC_THREAD load commands 6141 for (const auto &LC_THREAD_data : LC_THREAD_datas) { 6142 const size_t LC_THREAD_data_size = LC_THREAD_data.GetSize(); 6143 buffer.PutHex32(LC_THREAD); 6144 buffer.PutHex32(8 + LC_THREAD_data_size); // cmd + cmdsize + data 6145 buffer.Write(LC_THREAD_data.GetString().data(), 6146 LC_THREAD_data_size); 6147 } 6148 6149 // Write out all of the segment load commands 6150 for (const auto &segment : segment_load_commands) { 6151 printf("0x%8.8x 0x%8.8x [0x%16.16" PRIx64 " - 0x%16.16" PRIx64 6152 ") [0x%16.16" PRIx64 " 0x%16.16" PRIx64 6153 ") 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x]\n", 6154 segment.cmd, segment.cmdsize, segment.vmaddr, 6155 segment.vmaddr + segment.vmsize, segment.fileoff, 6156 segment.filesize, segment.maxprot, segment.initprot, 6157 segment.nsects, segment.flags); 6158 6159 buffer.PutHex32(segment.cmd); 6160 buffer.PutHex32(segment.cmdsize); 6161 buffer.PutRawBytes(segment.segname, sizeof(segment.segname)); 6162 if (addr_byte_size == 8) { 6163 buffer.PutHex64(segment.vmaddr); 6164 buffer.PutHex64(segment.vmsize); 6165 buffer.PutHex64(segment.fileoff); 6166 buffer.PutHex64(segment.filesize); 6167 } else { 6168 buffer.PutHex32(static_cast<uint32_t>(segment.vmaddr)); 6169 buffer.PutHex32(static_cast<uint32_t>(segment.vmsize)); 6170 buffer.PutHex32(static_cast<uint32_t>(segment.fileoff)); 6171 buffer.PutHex32(static_cast<uint32_t>(segment.filesize)); 6172 } 6173 buffer.PutHex32(segment.maxprot); 6174 buffer.PutHex32(segment.initprot); 6175 buffer.PutHex32(segment.nsects); 6176 buffer.PutHex32(segment.flags); 6177 } 6178 6179 File core_file; 6180 std::string core_file_path(outfile.GetPath()); 6181 error = core_file.Open(core_file_path.c_str(), 6182 File::eOpenOptionWrite | 6183 File::eOpenOptionTruncate | 6184 File::eOpenOptionCanCreate); 6185 if (error.Success()) { 6186 // Read 1 page at a time 6187 uint8_t bytes[0x1000]; 6188 // Write the mach header and load commands out to the core file 6189 size_t bytes_written = buffer.GetString().size(); 6190 error = core_file.Write(buffer.GetString().data(), bytes_written); 6191 if (error.Success()) { 6192 // Now write the file data for all memory segments in the process 6193 for (const auto &segment : segment_load_commands) { 6194 if (core_file.SeekFromStart(segment.fileoff) == -1) { 6195 error.SetErrorStringWithFormat( 6196 "unable to seek to offset 0x%" PRIx64 " in '%s'", 6197 segment.fileoff, core_file_path.c_str()); 6198 break; 6199 } 6200 6201 printf("Saving %" PRId64 6202 " bytes of data for memory region at 0x%" PRIx64 "\n", 6203 segment.vmsize, segment.vmaddr); 6204 addr_t bytes_left = segment.vmsize; 6205 addr_t addr = segment.vmaddr; 6206 Status memory_read_error; 6207 while (bytes_left > 0 && error.Success()) { 6208 const size_t bytes_to_read = 6209 bytes_left > sizeof(bytes) ? sizeof(bytes) : bytes_left; 6210 const size_t bytes_read = process_sp->ReadMemory( 6211 addr, bytes, bytes_to_read, memory_read_error); 6212 if (bytes_read == bytes_to_read) { 6213 size_t bytes_written = bytes_read; 6214 error = core_file.Write(bytes, bytes_written); 6215 bytes_left -= bytes_read; 6216 addr += bytes_read; 6217 } else { 6218 // Some pages within regions are not readable, those should 6219 // be zero filled 6220 memset(bytes, 0, bytes_to_read); 6221 size_t bytes_written = bytes_to_read; 6222 error = core_file.Write(bytes, bytes_written); 6223 bytes_left -= bytes_to_read; 6224 addr += bytes_to_read; 6225 } 6226 } 6227 } 6228 } 6229 } 6230 } else { 6231 error.SetErrorString( 6232 "process doesn't support getting memory region info"); 6233 } 6234 } 6235 return true; // This is the right plug to handle saving core files for 6236 // this process 6237 } 6238 } 6239 return false; 6240 } 6241